Deloitte · Primly Community

Deloitte machine learning engineer interview: what they asked, what mattered, what I'd do differently

ml_mike · 4 replies

Went through a Deloitte ML engineer loop in early 2026 for their AI and Data practice. Background: 8 years in ML, most recently recsys at a mid-size tech company. Here's the honest breakdown.

Process overview: Recruiter call (30 min) Take-home assessment (4-6 hours) Technical interview with two ML engineers (90 min) Client/stakeholder simulation interview (45 min) Offer

Take-home: They gave me a dataset and asked me to build a classification model, do EDA, explain my feature choices, and write up how I'd explain the model outputs to a non-technical client stakeholder. That last part is the important bit. They are not just hiring you to build models. They are hiring you to translate models into decisions for clients. The write-up probably mattered more than the model performance.

Technical interview: Heavy on ML fundamentals. They asked me to explain the bias-variance trade-off in plain language, walk through how I'd approach class imbalance in a real dataset, and explain when I'd choose gradient boosting over a neural network approach. For the neural network question they wanted practical reasoning about data size, interpretability requirements, and deployment constraints. Not a research interview. Very applied.

They also asked about MLOps: how I'd think about monitoring model drift in a client deployment, what tooling I've used, how I'd structure a re-training pipeline. This was significant. More emphasis on production reliability than I expected.

Client simulation round: This was unusual. They gave me a short scenario about a client whose fraud detection model was producing too many false positives and asked me to roleplay explaining the situation and options to the client. Pure communication exercise. No code. I've never seen this in an ML interview before but it makes sense for consulting.

What I'd do differently: Prep more on explaining ML outputs to business stakeholders. I was solid on the technical side but a little clunky on the translation layer. Also: know LLM fine-tuning basics because they are increasingly getting asked to stand up generative AI systems for enterprise clients.

Comp: my offer for senior ML engineer in the Chicago office was in the $155-165k base range plus project bonuses. Not FAANG but solid for consulting.

4 replies

ds_dmitri

The client communication round is real and I've heard about it from a few people who interviewed at the Big 4 firms. The thing is, in a product company your audience is internal. In consulting, you're explaining your model to someone who has 20 minutes and a P&L to protect. Totally different communication muscle.

market_realist

$155-165k base in Chicago as a senior ML engineer. Is that the typical band or did you negotiate it up? I'm curious how much room there is in the Deloitte offer process.

ml_mike

I negotiated. Initial offer was $148k. I came back with a competing offer data point (not a real competing offer, just market data) and they moved to $162k. The recruiter said the band was $140-170k for that level, so I was near the top. I think there's room if you have a clear ask and a rationale.

infra_ines

The MLOps section is good to hear about. A lot of companies say they want ML but actually want someone who can set up the serving infrastructure. Sounds like Deloitte at least knows the difference.