Communicating and Asking Questions with Seniors Who Lack Strong Technical Skills
I’ve gone through Taro's courses on communication and asking questions, which were insightful, especially in showing how juniors can learn from senior engineers with deep expertise. However, as an MLE, I face different challenges and would appreciate some insights or thoughts from Alex or Rahul.
For context, I majored in math and hold an MS in ML from a top 3 university, so I'd say I have a solid grasp of both the mathematical and practical aspects of ML. Machine Learning can be deeply mathematical, often requiring formal training to fully get it. Some SWEs who transition into MLE roles may lack this foundation, and while they might have been excellent SWEs, the gap in math knowledge can hinder their abilities as MLEs.
Here are a few challenges I face with some SWE-to-MLE seniors:
- Sometimes, my tech lead asks questions that suggest a lack of understanding of ML basics. While directly correcting them doesn’t seem right, what’s the best way to handle this?
- This becomes a bigger issue when they set project goals that are mathematically infeasible. Without a strong math background, they rely on intuition, making it hard to guide them away from misguided directions.
- Their informal grasp of ML can also complicate design documents by introducing unnecessary features that don’t align with the project’s objectives.
- As seniors, they can get defensive and dismiss alternative ideas, which is frustrating, especially when I would need to work on a project with a high risk of failing.
Discussion
(1 comment)- 1a month agoTech Lead/Manager at Meta, Pinterest, Kosei
To address this, you should view your job as one focused on education. Instead of correcting senior engineers in a one-off way, I'd document various examples where the lack of ML knowledge had a negative impact on the team.
Then put together a presentation about what you've observed, along with how to fix. I imagine the fix could either be:
- Address some of the gaps and talk about best practices (if the scope of the misunderstanding is limited)
- Create a brownbag series where you/others share some knowledge about ML (if there are many gaps and/or you don't have all the answers)
In your presentation, I'd reinforce that you care that the team is working on high-impact projects on the right timeline. Don't call out the background of various people on the team or specific people who don't have an adequate understanding of ML -- this wouldn't be received well.
Instead, focus on the patterns you've observed and the damage caused. (focus less on people, more on impact)
