Basis Machine Learning Engineer Interview Experience — Behavioral Plus Product Sense, Then an Agent Debug Round, Out Before Onsite

TL;DR: Basis process: a headhunter reached out, then a recruiter call, then two interview rounds (remote), and I got cut before the onsite. Round 1: behavioral with a deep dive into my resume, plus product sense / DFS-BFS, all in the same round. Round 2: debugging an agent codebase, AI allowed. I got no feedback, so what follows is purely my own recap.

Recruiter call (about 20-30 min): A headhunter contacted me first and explained the role and the rough process. After that I had a call with a recruiter from Basis. It was mainly:

• Self-introduction, my current status, why I'm looking at opportunities
• Fit for the role, the team, the interview process
• Expected salary and timeline. My advice is to prepare a 1-minute self-intro, your motivation, and your salary expectations in advance.

Round 1, Behavioral + Technical (60 min): It was one session, behavioral/resume in the first half, and then straight into a problem in the second half.

Part 1, Behavioral: I started with a self-introduction, and after that it was basically back-and-forth questions about my resume projects the whole way through. We also talked about company/role-related questions and the questions I had for the interviewer. They were all pretty standard questions, just asking about the experience on my resume.

There was also something like: design a product that uses AI, where the interviewer plays the PM and I'm the engineer, and then I talk with him about how to implement it.

Part 2, Technical: product sense + graph search. The problem was a product scenario. You first have to break down the requirements, then realize that it's essentially a graph search, and then choose DFS/BFS.

Round 2, Technical: debug an agent codebase (AI allowed): They gave me the codebase of an agent project and asked me to find the bugs and fix them, with AI assistance allowed. The problem itself isn't hard. It's more about code reading, locating the problem, and verifying the fix. Basically you need to look for what bugs there are in the agent tools or in the code.