There is no apples to apples comparison, FAANG just gives a ridiculous guaranteed amount of money... no matter how many different ways you fudge the numbers.

TC doesn't mean jack in startup world. You're always taking an outsized risk by joining a startup that may/may not IPO. By default, all non-public company shares are inflated, since there's no liquidity in them. Cash salary is the only part of TC you can depend on there.

Think of it this way: joining FAANG is like investing in the S&P 500, you'll get a decent return for the effort/time you put in.

Joining a startup is like investing in Gamestop. It could go to the moon or go boom.

AI/ML Eng @ Series C startup

Elliot Kang

i work 100% remote and have less than 1 YOE in software engineering

Engineering team was a similar size at my last job, seed stage. The general expectation was 3-4 months.

For mocks, you can use Figma if you need super high fidelity mocks, but that AI tool Alex mentioned should be enough.

For design docs, I found that either Notion or Google Docs in pageless format works really well.\
\
This gem should be a super-course:\
<https://www.jointaro.com/course/frontend-system-design-masterclass-building-playlists/>\
\
In the `Defining Requirements` section, there's a link to a design doc for a 1-week long feature build from Alex.

Something a mentor told me is that even if you're looking for a new job, you're still getting paid to work at a minimum 9-5.

idk if this helps, but it helped me a ton

It's a skill issue. No matter how "bad" the economy gets, there are going to be people who land tons of interviews. Your job is to be one of those people.

You can still land interviews as a new grad. I even dropped out of my Master's in CS and landed 3-5 interviews a week in Fall 2023, when the economy was supposedly terrible.

There will always be a market for exceptional talent. So show that you're exceptional.

As they say at Taro:
- get good
- land interviews
- pass interviews

MS doesn't count as 2 YOE in Big Tech/FAANG. People don't count it at startups either.

I would just apply anyways. The "shot" only counts if you make it to the interview loop.

Heck even at seed stage... most of the time, the existing code wouldn't have handled all the common cases. Then when you say "Refactor this code for me", Claude slashes up your code even more.

Conventional wisdom just doesn't work. You'll have to decide for yourself whether you want the accelerated growth of a startup life or not.

Big tech **on average** produces better results. But you are not the average (or so I hope). Remember these folks did really really well at FAANG compared to the average actually. They're telling you what worked for them.

From the information in your post, I can smell you've already made up your mind about the startup. Just do it.

Having grown up in the Bay Area, San Jose **does not equal** SF. You'll be driving up 1 hour to SF, which is doable but very painful.

**Career**

The AI appeal is really nice, but you'll seriously have to hustle to even have a chance. The competition is pretty intense. Startup scene is cool but it can get old when people are talking about Claude in a nightclub at 1am.

AI events are fun, there's a ton going on every day. They're okay for networking with fellow engineers. Founders usually speak at these events or work out of Shack15. It's not too hard to get a referral to Shack as of now, but that could change if they initiate a waitlist.

**Social Life**

I've been in a long-term relationship for 5 years, but my single friends tell me dating life is atrocious in SF/San Jose. The colloquial term for San Jose is "Man Jose".

Just leave things as is. You don't benefit at all from disclosing more information. Water under the bridge

Panicking in this situation is not going to do anything. Whenever you feel overwhelmed, step away from the computer and take a deep breath.

Something that concerns me is that you're instinctively falling into work anti-patterns.

1. Multi-tasking - Line up all your tasks in order of priority. Break down each task down so much that even a LLM could code up each atomic mini-task. Block 4-8 hour blocks of time to only do coding. Do the tasks one by one. Context switching is the ultimate killer of productivity.

2. Ask for Help - If your teammates are saying that you ask questions about trivial things, ask them the deeper question of how you can find these things on your own before asking next time. If they don't want to provide that advice, make sure to spend 15-20 minutes finding your own solution *and then* ask for help.

3. Work Longer Hours - Working longer hours only helps (barely) boost what's already working. In your situation, nothing's actually working so you need to revisit your work habits before you even do this.\
   \
   Go through Alex's course on "Junior to Mid Level Engineer" course. It singlehandedly helped me turn around my performance at my previous job. I know it seems like the world is crashing down, but sometimes you need to regroup and really make a strategy for moving forward.

<https://www.jointaro.com/course/grow-from-junior-to-mid-level-engineer-l3-to-l4/how-this-course-works/>

Morgan Wallen. There's nothing more encouraging than some slow, sad country music.

Don't have enough context to answer. What's the size of the company? Is this at a tech-first company or a company where tech is a support function?

Otherwise, we don't know the severity of several senior engineers leaving.

There's some good content here:\
\
<https://www.jointaro.com/question/YqrpDDCPbsGpg0lwrKXY/looking-to-transition-to-ml-engineer-what-do-you-want-to-know/>

I can't pick your niche for you, but here are some good constraints I wish I knew. Pick something that

- won't require you to make a top-tier ML conference paper
- has a mid-size open-source library (5-15k Github stars) that you can be a Top 3 contributor in within 3-6 months
- is actually interesting to you. You don't need to already be a user, but you should have an actual interest in playing around with adjacent tools

Hope this saves you time

\*\*I wrote this in case you pursue ML. Although you did mention "data roles"

As a software engineer, you have the power to will things into existence (with some additional typing of course). You don't need permission to build anything.

PM is very often a trap for software engineers, and it's really hard to go back to SWE once you've moved to the dark side. As Rahul hinted at, a lot of big tech PMs basically just push metrics and fill out Jira tickets.

Why not just focus on being really, really good? Opportunities to collaborate with exceptional people will open up as you do.

The lightsaber is only as strong as the wielder.

Just turn off all AI tools and use Google, read the docs yourself, and build stuff.

AI hallucinates convincingly, so it's hard to actually increase your productivity with AI unless you take a nuanced approach. You still need to do your homework on specifics.

You're not alone. I was jamming my keyboard begging ChatGPT to build me a multi-layered Dockerfile and it just kept spitting garbage 🫠

Navigating a somewhat adjacent situation myself, you're not alone 😓

now that i think about it, my answer was very un-nuanced. you can ping me on linkedin if you have specific questions, I ran the largest AI meetup in the bay area for a year

Try cerebralvalley.ai


- hackathons at AGI House in Hillsborough (not the SF one) are pretty fun, top tier engineers show up. Andrew Ng was squinting really hard at my failed demo one time 😭

not searching rn. but yeah. I found that all mainstream social media has garbage sponsored content as 90% of the feed, making something with less sponsored stuff is awesome

Tbh, you're probably not making progress on your DSA/interview prep because you live in an attic and have no coworkers to talk to. Your environment outside of work often affects your work deeply. So maybe instead of $400K being your immediate goal, just get out of your parents' place.

Personal experience: I dropped out of my Master's and lived at home for a year before finding a SWE job. It was a terrible existence. I still do live at home but have a job now so it's much more bearable.


Sorry don't have any inside info. But you can find a lot of this info out yourself if you can secure the interview.

The best way to find out is to apply now and secure the interview. Then you can ask the hiring team what the interview process looks like + some detailed follow-up questions. If you're underprepared, respond slower to Elastic and push back the interviews so you can prepare

This is gonna ruffle some feathers, but I think that the best way to prep for technical interviews is just to do more technical interviews. The nerves are the worst part for most people. You can farm early/mid stage startups for technical interviews tbh (if you already generate tons of interviews)

Sounds like an awesome idea. Even better, if you find any alumni from your school's student clubs that you are involved in, that can make the connection even better

Just leave. There's plenty of other AI startups with funding. It looks like your team couldn't find customers before the lights went out. That's too bad. Dust yourself off and go find something else to do

Whenever someone asks you if everything is clear, just pause for a few seconds and think through whether you can visualize the implementation clearly (lines of code). The silence may be awkward, but it'll save you lots of headaches later.

^ what Rahul said.

Now that I think about it, this is also how you proactively get good at contributing to open-source.

There's some great input here. So why does school matter? (at least initially)

People are always looking for signs of exceptional ability. And at the beginning of your journey, school is often the only indicator of ability if you have zero internships or cool projects.

Top company is an even better filter than school because it's a more recent indicator of ability. Even at top schools, very few people actually end up succeeding as much as you think (FAANG, IB, etc)

Ah I see. Then yeah maybe it would be better to work under someone with much more experience for a bit. You can always calibrate your work-life balance up or down as you progress.

I'm also on the extreme WLB end right now, went on 3 trips in the past 2 months aha

If it's generative AI, I'd wait until they're slightly later like Series B or C. There's so many knockoff GenAI companies that they're extremely unpredictable right now. 

YC-backed doesn't mean anything these days, as YC is now shovelling out AI startups like there's no tomorrow. If you don't plan to stay longer than a year, your shares won't vest anyway so I would just go somewhere else with the plan to stay at least 1-2 years.

I wouldn't trade the free time, as you could be sharpening your skills on your own time and not someone else's. That said, it requires more discipline than when you're a busy employee.

and you can probably do that in the "tell me about yourself/background" blah blah blah

Now that I'm re-reading this, "cleaver of truth" sounds really corny. The point is, you need to proactively address any concerns your interviewer may have before they bring it up 🫣

It's a trading firm. Compensation should not be a taboo subject. I understand that many people try to dance around touchy subjects, but that only raises more alarm bells in the interviewer's mind. And compensation is only a lagging indicator of organizational problems, so feel free to add 1 or 2 more of your complaints when people ask. (gracefully, of course 😅)

When someone has doubts about you, your responsibility is to cut through that doubt with the cleaver of truth.

I started interviewing for new roles 4 months into my previous job. The first thing interviewers asked was why I was leaving so early. So when I noticed the trend, I pre-emptively told them why I was leaving my current job (lots of turnover, etc). They never asked again.

Everyone has to play the interview game. But you get to decide how you'll play it. Are you going to dodge all the real questions until you get eliminated or are you going to proactively attack the interviews and get the damn offer?

That's up to you

Not informed enough to know whether this works for higher levels, but this is a good strategy for now at least. Erring on the side of overcommunication is better, especially if you're fully/partially remote.

Don't be discouraged if your boss doesn't respond, he/she will still read those updates and have some piece of mind. Plus if you're getting blocked on a 5-day task, you'll be able to catch it Tuesday and not the following Monday

Oh. and. Try to carve out 4-6 hour blocks of time where you're doing some deep work coding. Alex mentions this in his ["Maximize your productivity"](https://www.jointaro.com/course/maximize-your-productivity-as-a-software-engineer/) course, and that really helps from personal experience

Like Sai and Charlie mentioned, it takes a certain amount of effort to make AI actually work for you when coding. AI is good for:

- extremely common code patterns
- shuffling simple stuff around (variable name changes, small function refactors)

Larger design decisions that span multiple functions/scripts don't work very well. Similar to how Alex mentions breaking down units of work, breaking down work even more with atomic tasks helps you get a way better ROI on AI tools. I use this process:

- break down tasks to the function level
- insert some comments to annotate what I want Cursor to do
- use the chat interface and feed in context for each mini-task
- audit your code by looking at it/testing it

Rinse and repeat.

The way I like to think about it is that AI is 100% useful on 20% of your work as a SWE (doing mundane stuff like regex, etc). Cursor can't help you distill requirements or make a great design. Cursor also won't help you manage expectations with your boss.

To add onto this: just don't do any of this work on a company laptop if you plan to commercialize

If you have the flexibility to, travel. I'm on my third trip in 3 months and it's been so helpful in resetting + building new work habits.

It can be difficult to break out of bad habits sometimes, so physically relocating even for a few days can make your brain flow differently

Both work. However you need to answer this question first: why should this person hire you out of hundreds, probably thousands of applicants?

**Why** is so so important. And the way that your "why" manifests is by doing deep work on relevant projects to the CS subfield you want to be in. It took me the longest time to put this idea together mentally

Hm. While your intentions are good, I think you're leading too much with "look at me I found a problem in your system". Could you try just reaching out with the premise of wanting to learn more about the company/team? Then you can lean more into "I built something and found out some cool stuff about Netflix' production app this way" as the conversation warms up.

Open-source really is king when it comes to this kind of "attention grabbing". Side projects can be hit-or-miss unless you're cracked.

Awesome that you're switching to SWE! Piggybacking off of Alex:

- For non-frontend roles, I don't think they'll ask you too much framework-specific questions
- Leaning into your strengths as someone who works with data-intensive applications will help a ton
- As Alex said, in practice TDD can get in the way of innovation. You can just follow your new team's testing culture when you join

Finally, on knocking down those doors to get those interviews:
- If your project is somewhat similar to what the team is looking for, you can ping the hiring manager, or on the application submit your project code as a work sample. I've gotten several interviews this way, and also gotten people who were on the fence about me to at least move me to the next round of interviews

Like Rahul and Charlie said, I think it's a good sign that you're still hanging in there. All things considered, your case is on the better side given all the layoffs/turbulence over the past several years.

I graduated college in 2019 and some of my peers never recovered

no, this is just a measure of how big of an AGI House fanboy/girl you are 😆

You have no idea how much pain running a startup can bring. It really can be like having broken glass for breakfast every single morning.

I think the premise of your questions should be different. Life is a marathon. Instead of jumping the gun and trying to learn everything at once, why not learn aggressively at a sustainable rate? Most of the time, the way to do this is to get a hybrid/on-site SWE job and moonlight on the side. Then, you can go 100% remote for even more freedom. Then, you can quit your job once the startup makes $10-20k USD/month.

I know you have this patience: you stuck around for a CS degree. But going 5 years at full speed doesn't do you any good if it requires 2 years of therapy to recover.

To add to Alex's statement, it's easy to misread data. For example, those metrics like MMLU don't mean much when you're trying to make these "AI agents".

Benchmarks, by design, look really good. But AI projects/papers often solve a really specific problem, so blindly hoping research results generalize to your use case doesn't work.

When learning any science, the real skill is finding which information is relevant/actionable to you. It's 99.9% noise out there in GenAI. You need to do side projects/implement papers yourself to get any real alpha in GenAI tbh.

I'm also curious about this question. Currently, I'm writing some Terraform/Chef to spin up resources to run distributed training jobs on the cloud.

We don't all have a 10000 GPU megacluster to run pretraining on 😿

ooh. I'd take the Anduril SWE option. Not because your stocks will hit the moon, but because you'll stay a software engineer. 

As a business engineer, you wouldn't do as much hardcore coding.

**take this with a grain of salt though, some more informed people will have better opinions. But one thing is for sure, please don't count on the non-cash portion of TC at a non-public company (you'll only end up disappointed)

Low. Kaggle ranking is just a number. Open-source and building in public generates far more inbound than Kaggle. You can generate dozens of inbound interviews this way. Just do cool stuff, then write a little about it. Rinse and repeat.

The reason for this is **people want to see cool shit**. A ranking number is boring.

To answer all your immediate questions:
Papers are a good way, but they take lots of time and your work is unlikely to be rewarded. There is no certainty that you'll publish a decent paper, if at all. Generally ML people don't care about applied ML papers as much. Worthwhile top tier papers take 12-18 months from idea inception to publication acceptance/release **at a minimum**. Second tier conferences like AAAI and KD are still good, but less prestigious. Workshop papers are in much lower regard to real papers.

Some general points:
- ROI on papers is low, given how much time/effort you're putting in. If you want to do purely model research though, this might be the only way
- Very few ML engineers are purely doing model work. The majority are doing systems for ML, SWE for ML, or infra for ML.
- open-source is pretty good ROI if you know what you're doing, there's a Taro course on this
- Switching from engineering work to science work has it's own risks. It's hard to measure research productivity and you might find yourself at a C-tier research shop with some bums

The main point is to always line something up as you never want to be in that OPT phase. You can do this! 💪

Alex's advice is spot on. Being proactive/working well with others is a good defense in itself. There are ways to be proactive about your fate without playing full-on offense, such as covering all your bases like Alex mentioned

You're definitely not alone in disliking this manager person. And you're not the crazy one (in case anyone tries to gaslight you)

Given your visa situation, I think it'd be good to stick it out for some time at Amazon. Your number one goal should be to buy time as someone on a work visa. It's like playing poker or trading stocks: never get completely wiped out. If your visa expires, it'll be way harder to find another job in the states.

IMO visa >>>> compensation right now. If you find another company willing to sponsor you though,  then yes the visa situation is taken care of and then you can get greedy on compensation.

**In case others have further input, how bad is the compensation difference? I think it's okay if your comp is lower than the "average" people on Levels.fyi

Hm. I graduated at the same time too. With all the data points you've given, I don't think this is even a tech stack issue. You're **using bootcamps/degrees as an intellectual crutch**.

Like,

- more bootcamps/degrees won't help you feel smarter
- switching tech stacks won't help you build cooler things

You still need the core SWE job search things like writing a great resume and passing the interview rounds. But doing the same thing everyone else has been doing... phoning in for a 9-5 isn't going to cut it anymore.

If I were you, I would

- Find a meaningful [open-source](https://www.jointaro.com/course/become-an-open-source-master/) project that you'd be willing to commit to every single day for \~2 months
- Write a killer resume. If you're not getting any callbacks after 15-20 applications, it's a dud. **Every line should be a banger**.
- Practice interviewing, once you start landing at least 2-3 a week. Your resume isn't an effective hook until then.

You can do it. If you need some help, we can set up a much more granular action plan to get this done. You're in a much better situation than I was in mid 2023 (MS dropout, no CS degree). Ping me on Slack or Linkedin if you need to

### Personal Anecdote

You're in a bit of a hole, but that's fine. I've been in a similar situation before. It's never too late to turn it around. My brutally-honest resume is:

- Graduated 2019, bombed maybe 30-40 interviews (including Confluent when they were $1B valuation)
- 2020, got wrecked by pandemic + worked some odd jobs
- 2021 - 2022, played mobile games on my iPad + did CS courses
- 2021-2023, went through the motions doing a Master's in CS
- 2023, dropped out and did open-source
- 2024, finally got an AI job after I got tired of working in a restaurant (tips were good though)

Not the optimal path. But hey, at least I have fun building things now. And I washed dishes really fast after Thanksgiving dinner.

lol. So one of my friends had a similar interview experience when he did his onsite at Amazon. It was two questions and they just chatted the rest of the time. He got an offer and ended up being a really high performer at Amazon.

But back to your situation. The question behind your initial question is whether your super short onsite is a sign you failed the interview.

You're overthinking after the interview. You already did everything you possibly could do. They could have put you through a grueling interview loop if they wanted to. **If they wanted to, they would**.

TLDR; you're overthinking, just wait it out.

I don't think you should have any guilt regarding "I'm getting paid more now, therefore I should work more hours".

#### Getting Things Done on Time

That being said, your projects need to get (mostly) done on time. Maybe you could discuss with your manager why your projects are running behind. A couple common issues are:

1. Scope creep. A 3-day issue turned into a 7-day issue because you or someone else added to it. **Fix**: Make a follow-up ticket/diff for the new additions.
2. Getting rug-pulled. The team agreed to the scope of a ticket/project but they say they wanted way more features at the finish line. **Fix**: Get people to agree to written requirements before implementation.
3. Code review takes too long. Your code is too annoying to review, so nobody does it. **Fix**: Alex discusses splitting diffs in the [Level Up Your Code Quality](https://www.jointaro.com/course/level-up-your-code-quality-as-a-software-engineer/) course

Don't be afraid to call people out on (1) and (2), as it's now your reputation on the line.

#### Gracefully Handling Lateness

Regarding being late for projects, things always come up in a startup. By default, some projects are going to be late. It's how you handle the lateness I think, that really matters. Some key things to ask yourself are:

- Are you communicating with the team that solving issue A will delay project B by # days? How well is news like this received?
- Is your dev pace matching/exceeding current members of the team?
- Who is doing your performance review?

#### Performance Shouldn't Be a Surprise

Finally, your performance review is basically at the mercy of whoever is in charge of that. I think that given you're not even 90 days in, it's impossible to give a definitive or close-to-precise answer. If someone got let go in 90 days, that means they legitimately did nothing, so I wouldn't worry too much about that. (if that's only a single data point)

To answer your question, **there is never a guarantee that your team will let you know that you're underperforming**. If I were in your shoes, I'd bring up your concerns with your manager (who I'm guessing will do your performance review) and see where you guys can iron things out. If possible, it's good to gather your thoughts and do this asap.

Hi, seems like you're on the right path! I understand your urgency, as it's a 6-month contract and you're trying to secure a full-time gig.

I found these courses really helpful at my last job where I had to deal with some serious leadership drama, switch teams, and turn around my performance as a "Research Engineer"-turned "SWE with a sprinkle of AI".

### Courses

#### Junior to Mid Level Engineer

#### <https://www.jointaro.com/course/grow-from-junior-to-mid-level-engineer-l3-to-l4/>

Helps with code quality, communication, and distilling requirements for projects.

#### Leveling up Code Quality

#### <https://www.jointaro.com/course/level-up-your-code-quality-as-a-software-engineer/>

Really helps you get into the feedback loop of improving code quality with the more senior engineers, as they'll get "setting the bar" for what gets pushed into the codebase and what's rejected.

#### Performance Optimizations, for work

#### <https://www.jointaro.com/course/maximize-your-productivity-as-a-software-engineer/>

If you feel like you could be way more productive at work, you're not wrong. It's an iterative process, but there's some good pointers here. Alex also politely tells you not to be a lazy bum.

\*\*Doing these courses in sequence yielded really good results. It takes a few days to go through all of these, and maybe 2-3 months to really internalize.

### Masterclasses

#### Shipping Real Features

#### <https://www.jointaro.com/course/frontend-system-design-masterclass-building-playlists/>

I borrowed the design doc format for this masterclass and got amazing feedback from my team. There's three ways to use this IMO:

If you're at a larger company, hopefully they already have a uniform design doc format and you can just take away some nuggets of wisdom from this class. If you're at a startup, chances are you aren't getting much guidance, so you'll have to bootstrap a format on your own and iterate on it. You can also just practice by **writing design docs for your side projects**. Here's a somewhat incomplete design doc I wrote for making a [Figma AI copilot](https://docs.google.com/document/d/16hKvPpoF2Q8fyJd1qnvhHtuRPMJKkfiYIiWvmBcTC4s/edit?usp=sharing)

These can be fun to write in your free time + you get to really think through your ideas.

Anyway, hope this helps.

Wow you write well! I really like your point about being proactive finding work to do. This definitely makes life easier, as you're no longer at the mercy of your Jira board

In my enthusiasm, it looks like I missed a few points 😅 Some things I notice about really, really good engineers is that

#### Breaking Down Work
They're just so good at breaking things into atomic tasks. If there's a whole new workflow, they break it down into as many trivial tasks as possible and knock them out.

Some of Alex's videos on "how real system design/feature planning is done" are super helpful. There's like 3-5 of these at least somewhere.

#### Getting the right people involved
If there's a feature that involves approval from 3+ people, I've seen them just get ahead of the work and gather requirements super efficiently.

TLDR; good engineers are really good at **doing the preparation before coding** so they have to write 70% less code when the implementation actually begins. WLB does exist, but I think you'll have to fight hard for it.(being efficient, constantly improving, picking the right companies/teams)

Oh man. So. I said yes to everything when I first started being a SWE 6 months ago. It was a huge mistake. I was trying to do too many things at once. Ultimately, I took a huge reputation hit and it took months to recover.

So after reading your post several times, I'm getting the impression that you're putting too much pressure on yourself. The expectation of any new employee is to **get the most important work done first**. That's pretty much it. Then you add stuff like code review, refactoring, etc.

I wish I knew a few of these non-ENG related (but even more important) things before I got into tech... 😴

#### Having Boundaries
If your teammates ask you for help and you're swamped with work, just say no. They'll understand, almost any reasonable person will. This is really important because if you never say no, people may just keep asking you to do things.

#### Stretching Deadlines (in moderation)
You're never going to have 100% days all the time. So if there's a seemingly small project but there's a lengthy review process, it's good to give yourself 20-30% breathing room to put less pressure on yourself.

#### Not trying to learn everything at once
If you try to learn everything at once, you'll learn nothing. How are you supposed to know what the code reviewers on your team look for, before you've even landed any code? 

It takes time to meld your coding style with the team's. It takes time to give consistent and thoughtful code reviews to your coworkers. It takes time to add value via ad-hoc tasks.

^side note: I think it's actually harder to add value with ad-hoc tasks than your main work. They are prone to be poorly scoped and not prioritized properly, so it could be a huge time sink to work on these if you're new.

#### Final Thoughts
Enjoy the grind! Make sure to have some fun. Touch some grass. Blow some money on concert tickets or something. As a highly paid SWE, it's your turn to stimulate the economy 💸

Entry-Level Software Engineer at Seed Startup

There should be resources online on the expected equity percentage. That being said, I'd give the "equity" at least a 70% haircut when you tack it onto base as "TC".

Talking to the engineers would definitely help.

Oh yeah, final thing. If it's a Series A startup, ask what the milestones are to hit Series B and where the team is at in regards to that. If the startup is at $0.1 million ARR and it needs to hit $20 million, that's just an uphill battle.

Your concerns are valid. There is only so much you can find out about a startup during the interview process. And it's unlikely they'll tell you about any of the real problems, so you have a ton of guess work to do. If they do tell you about real struggles, that's a pretty good green flag.

Founders tend to say whatever they need to in order to get early employees to join. Sadly, you don't have Glassdoor at this stage to do some due diligence.

### Common Red Flags

#### Shiny Demo

Never, ever believe any demo that they give you (especially if it's an AI demo). It's most likely fake, unless they let you use the tool yourself.

#### Equity-Heavy offer

Early stage startup valuations are rather fuzzy. And if they have stacked SAFE contracts, your equity can get diluted really fast. Don't let startup founders give you this "but your equity is worth $5 million right now according to my personal valuation...", that's just a lie.

#### High Turnover

Look up both current and former employees on Linkedin. If the turnover is really high, chances are you're looking at some unstable personalities steering the ship.

#### Nobody Leveling Up at the Startup

Ask how people get promoted or significant increases in comp. If nobody has gotten increased responsibilities/comp in the past 6 months/year, that means nobody is given the opportunity to grow.

#### Low-No revenue/High Churn

Money talks. If a startup isn't getting traction, it's not getting traction.

### TLDR;

Sadly, I can't name all the red flags. There are many unscrupulous founders who will bullshit you to the moon and back. But by asking hard questions during the interview process, hopefully you'll be able to filter out most of the grifters.

## Addressing Resume Gaps
You're not alone, I have a gap on my resume too. Did a Master's in CS and then dropped out to pursue this GenAI thing. So yeah, I get it.

### The Market and Resume Gaps
I got an offer in April this year, so I've experienced some of this firsthand. Given how hard the tech market has been since last year, I doubt too many recruiters will throw out your CV because of a resume gap. And if they do, they don't deserve you anyways (as long as you've been doing something to progress your goals)

### Attitude During Hard Times
I think the more foundational thing is that you don't give up. Everyone gets stuck at some point, whether it's not getting into a top tier school, hating their first job, or not getting promoted at their first job for several years. I even had friends get into medical school after 4 years of trying. 

If you *really* want to do this and don't see yourself doing anything else (at least in the 3-5 year timeframe)... just stick with it. 



It seems like you're wondering what creating a startup will be like. 

Here's a quick way to see what it's like, without quitting your job and making your partner angry:

Just go live a day in the life with a startup founder (preferably several to get a representative view).

If you work remote 1-2 days a week, you can go to an AI event or hackathon and look for startup founders to cowork with. Then you can pick their brain over lunch/dinner after coworking at a cafe or something. (they might even invite you to a founder hotbed like SHACK15, which has even more founders)

Early stage founders tend to have much more insight into things like:
- what the funding climate is actually like in your region
- how to deal with the stress of being an early stage founder
- hiring the first few people, etc

This is important because in 2024, the ground truth and what mainstream news reports about tech diverges significantly at times.

Anyway, hope this helps. I haven't started a startup yet, but it gives me peace to at least know what I'd be signing up for if I did.



I think most of those startups are definitely later stage, which are much safer. It's just that there is the constant risk of being no longer seen as that "breakout startup". For example, OpenAI releasing a killer search engine would hurt Perplexity.

Mistral this time last year would have been considered a breakout startup :/

I'd just be wary of any of the generative AI startups on that list, some of them have serious customer churn (so they wouldn't be considered breakout startups) and are having trouble raising a next funding round.


Well. To give you some hope, I was jamming Cursor AI with a bunch of questions for 20 minutes about how the opentelemetry library worked. What I really needed was a 3-line change, which I would have known had I worked with the library before.

Cursor definitely helps. But I've noticed that the larger the codebase, the worse Cursor does. For max accuracy, I've noticed that it does best on knowable work that has lots of documentation.

Things Cursor is bad at:
- refactoring code where the bugs/edge cases are unknown
- writing code that based on other internal code

btw I'm not sure if o1 actually helps, it's a little too expensive for my taste. I just stick to o1-mini for now. But hey, I'll take the free productivity boost for $20/month

Well if you can get into OpenAI or Anthropic, absolutely. But if you're part of the 99.99% that can't get in, it's okay. I knew someone who got into Midjourney and he had to be so so proactive to even get noticed.

Taro can give you all the resources in the world, but the harsh reality is that there is very little outside information on private companies. So if I were in your shoes, I'd try for big tech simply because there is much less unnecessary risk.

You never know if your startup
- is about to hit a major down round
- doesn't have product-market fit
- has 2 weeks of payroll left
- even has a good engineering culture

Especially if you're looking for a quick cashout, founders will totally sell you that dream to get you to join.

Some harsh truths:
- Work-life balance is worse for sure, unlike FAANG there are tighter deadlines because you should have shipped yesterday
- TC doesn't mean anything for you, it's only for your employer to rationalize paying you less
- Chances are, nobody has heard of your startup. And 99% of the time they never will.

If you think this would never happen to a break-out startup, it can at any moment. Meta is absolutely killing every other open-source LLM startup, including Mistral (which had the world's best cost/performance LLM for a few months).

Now, reasons to work at a startup despite all the instability:
- you get to ship fast, features take 1 week instead of 4 months
- you can impact the startup's success
- you get to work on a variety of things (infra, features, maybe even some ML)
- your coworkers are likely to enjoy their job
- unlike big tech, they won't sunset your project after you've sweat blood finishing it

Ultimately, nothing from the startup will likely remain. The equity can be worth pennies, nobody will remember how many all-nighters you pulled, and people don't care about the startups that never made it. What remains is your skills and experiences.

Well, at least that's what I tell myself so I can sleep at night

I was in your same position exactly a year ago, so this hits really close to home. It was so hard to get interviews that even with spam applying, I only got 2 callbacks in a month. Then I started contributing to open-source, and things started looking up. Here's some pointers:

\*\*For those who are later in their career, I think these principles still apply. The impact of doing this earlier in your career has more of a 0-1 effect than the 1-5 effect someone later in their career would see IMO

### **Pick A Niche**

So there's a deluge of MSCS graduates who want to work in AI. You need to focus on a single direction. There's several routes you can take:

- Model Serving
- Classic ML model development (collaborative filtering, etc)
- Large ML models (LLMs, stable diffusion, etc)
- Fine-tuning
- Generative AI tools
- ML infra tools like Flyte/Airflow
- Classic ML model dev and Large Model work typically require top conference papers to be in the top 1%

Before you dive into any route, make sure you do some due diligence so that you know roughly what it takes to be the top 1% in anything. Then decide if you're willing to put in the work.

### **Pick The Project Size**

Then there's size of the open source project. Small libraries/frameworks are too unknown to get you any real attention with recruiters. Large libraries/frameworks will take 1-2 years to get on the PMC.

Small projects are things like util libraries. Large projects are established frameworks like PyTorch or Kubernetes.

The "goldilocks zone" IMO is medium-size projects that are rapidly gaining traction. They have enough things on their roadmap that'll add immediate value and they should be the most welcoming of new contributors.

### **Making Core Contributions**

Start with something small. No sane person would ever let a stranger add changes to their core codebase. Start with something small like a cherry-pick commit or an integration.

Then it's time to align your contributions to the actual product roadmap. Attend contributor meetings. Talk to other contributors on Slack/Discord. Get involved.

As you contribute, you'll see yourself go up the "top contributors" list. You'll see yourself writing better PRs, writing cleaner code, and working on more impactful features.

### **Spinning Open-source Into Opportunities**

Getting to the top 10 or top 5 contributors for a medium-size library doesn't take that long. Even in the span of 1-2 months, you can get to Top 5 if you put in full-time effort.

Now you have some awesome résumé bullet points:

- Built core features: X, Y, Z, etc
- Top 5 contributor for Github project with 20k stars and 500K downloads/month

Now go! Find a project and contribute. You'll learn a lot + be helping out the community. I think that you can only re-write your resume so many times before you need to make a fundamental change. Making open-source contributions can help you make that fundamental change to your profile.

Calling all ~~autobots~~ Bay Area Tarodactyls! Join fellow acolytes at The Peak of Fremont where you can test your ~~graph algorithm~~ skills and exercise your fingertips for peak WPM on that split keyboard!

Location:

[**4020 Technology Pl Suite 1, Fremont, CA 94538**](https://www.google.com/maps/search/?api=1&query=4020%20Technology%20Pl%20Suite%201%2C%20Fremont%2C%20CA%2094538)

Fees:

Climbing passes are \~$30

Folks who will be there:

- top tier software engineers
- the denizens of Taro Nation
- notSamAltman

Be there or be squared\
\
⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⠛⠛⠻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀⠀⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀⠀⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀⠀⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⠟⠋⠉⠻⣿

⣿⣿⣿⠁⠀⠈⢿⣿⣿⣿⣿⣇⠀⠀⠀⠻⣿⣿⣿⣿⣿⠟⠋⠁⠀⠀⠀⠀⢀⣿

⣿⣿⣿⣄⠀⠀⠀⠙⢿⣿⣿⣿⣆⠀⠀⠀⠹⣿⣿⡿⠁⠀⠀⠀⣀⣤⣴⣾⣿⣿

⣿⣿⣿⣿⣦⡀⠀⠀⠀⠹⣿⣿⣿⣦⠀⠀⠀⠘⣿⠁⠀⠀⢀⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⡄⠀⠀⠀⣿⣿⠟⠋⠀⠀⠀⠀⠈⠀⠀⢀⣾⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⠁⠀⠀⠀⠋⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⣼⣿⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣤⣾⣿⣿⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣿⣿⣿⣿⡀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣠⣾⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿

⣿⡿⠋⠁⠀⠈⠙⢆⠀⠀⠀⠀⠀⣠⣶⣿⣿⣿⡿⠛⠁⠀⠹⣿⣿⣿⣿⣿⣿⣿

⣿⠁⠀⠀⠀⠀⠀⠈⣧⠀⠀⠀⠀⠈⠙⠻⠟⠁⠀⠀⠀⠀⣠⣿⣿⣿⣿⣿⣿⣿

⣿⣄⠀⠀⠀⠀⠀⣰⣿⣿⣶⣤⣀⡀⠀⠀⠀⠀⠀⣠⣴⣿⣿⣿⣿⣿⣿⣿⣿⣿

⣿⣿⣷⣤⣤⣴⣾⣿⣿⣿⣿⣿⣿⣿⣷⣦⣤⣴⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⠀