In 2026 there have been almost no Amazon DS interview write-ups here on 1point3acres — I'm starting to wonder if they're on a hiring freeze. I'm writing one up so people have something to reference.
Around mid-March I had applied to 2 AS (Applied Scientist) roles. A recruiter reached out to me on LinkedIn and scheduled a recruiter call a few days later. On that call they told me AS roles are mostly filled by PhDs and are very competitive, and suggested I apply for this DS role instead (leaning LLM). A few days after I applied, I got a technical screen notice, and we scheduled the screen a week out. The recruiter was really good about it — they gave me a prep call and walked me through the likely questions and how to approach them in detail:
Top Priority
- Science breadth is super critical.
- Be ready to explain what you are working on, the particular problem, the tradeoffs, and why you made specific decisions.
Likely Science Breadth Topics
- LLM fundamentals
- Random forest
- Boosting
- Bagging
- Bias-variance tradeoff
Coding Guidance (SQL)
- Clarify first before solving.
- Explain why you are choosing a method.
- Talk through tradeoffs.
- Be explicit about data structures.
BQ
They said it would only test this one: Deliver Results. It can cover:
- Tight deadline
- Final outcome
- Large-scale project
- Multiple stakeholders
- Push back with data when needed
- Explain tradeoffs clearly
- Convince stakeholders
Here's what actually happened in the interview:
The interviewer introduced what the team does.
I gave my self-introduction.
BQ1: Tell me about one challenging project. I steered it toward tradeoffs, decisions, stakeholders, and results, but while answering I forgot to emphasize bias for action — that might be what got me dinged.
BQ2: Tell me what you've done differently. This one I had to come up with on the spot (I hadn't prepared enough for it, but overall I don't think I answered badly — this could also be what got me dinged).
Since all the stories I'd prepared were LLM-leaning, deliver-results stories, and this team also wants to hire for the LLM direction, the interviewer focused a lot on asking about the RAG system and some of its details. I feel like my answers were fine there.
The SQL question was a fairly simple filter + join + group by + sum — I used a CTE, no window function needed. Then, because I asked whether nulls should be handled, they dug in further and asked how to also count nulls into the same final table. Then they asked me to return the sum for the most recent year, and I asked whether they meant a specific calendar year or something else — the interviewer said that was actually a good question that people often overlook, and clarified they meant the trailing 365 days from the current date. Also pretty simple.
I asked the interviewer two of my usual questions. What does this role need to deliver in the next 1-2 quarters — they said it's roughly LLM-related goals, and I looped that back into how I happen to be strong in LLM and could onboard and deliver quickly. How much of this role is research — they said there would be some research work involved, and I looped that back into having research ability too.
The whole conversation felt pleasant. I think the reason I didn't pass was most likely that I wasn't prepared enough on the BQ side.
I got rejected the next day. Felt a little down about it, but overall I still think it came down to the BQ answers.
Hoping this helps anyone currently interviewing for DS roles.
Discussion
Loading comments…