Datadog Machine Learning Engineer interview questions

5 Machine Learning Engineers have shared notes from their Datadog interviews, with the questions they were asked in each round. That is 17% of the 29 Datadog interview notes on Screna AI.

5 Machine Learning Engineer interview notes

Datadog·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Datadog ML engineer interview threw a pretty gnarly systems question at me about fitting OLS on a dataset that won't fit in memory. The whole thing felt more like a design discussion than a coding screen, which I wasn't fully prepared for.
  • You have a 25 GB CSV file with over a million rows and 3,000 float features per row. It doesn't fit in memory. How do you fit an OLS regression on it? Walk through streaming aggregation, numerical stability, parallelization, and how you'd verify your result against a small baseline.

“This one took me a minute to even parse.”

View Post

Datadog·Machine Learning Engineer·Technical Phone Screen

May 2026
Technical screen for an ML Engineer role at Datadog focused entirely on implementing Grouped-Query Attention from scratch. Pretty deep dive into transformer internals, not the usual LeetCode grind.
  • Implement the forward pass of a Grouped-Query Attention module from scratch using only basic tensor operations, no high-level attention helpers.
  • How much memory does GQA save versus standard multi-head attention during autoregressive decoding, and why is memory bandwidth the bottleneck at decode time rather than raw compute?
  • How would you convert an existing multi-head attention checkpoint into GQA? Specifically, how do you initialize the G KV heads from the original H KV heads?
  • Where do rotary position embeddings get applied in a GQA setup, and does the grouped structure change anything about that?
  • Implement the incremental decode step for GQA: given a cached K and V of length t and one new token, produce the next output and update the cache. What shapes change?

“The core of the whole interview.”

View Post

Datadog·Machine Learning Engineer·Onsite - System Design / Architecture

May 2026
Datadog ML engineer interview that went deep into monitoring system design, specifically around alerting pipelines for web server latency. The question was meaty and I don't think I fully nailed the on-call policy portion, but the technical detection side felt okay.
  • Design a pipeline to monitor latency across dozens of web servers (each reporting average response time per second), including algorithms to detect degraded performance and a paging policy for on-call engineers that handles ack, snooze, and re-page intervals to reduce alert fatigue without missing real regressions.

“This is basically two questions stitched together and I treated them separately which I think hurt me.”

View Post

Datadog·Machine Learning Engineer·Technical Phone Screen

Apr 2026
ML engineer round at Datadog that was basically one big deep-dive into loss functions for imbalanced classification. The question sounds contained but the follow-ups kept branching and I wasn't ready for how far they wanted to go.
  • Implement focal loss from scratch for binary classification using only basic tensor operations. Your implementation must handle raw logits (not probabilities), support a focusing parameter gamma and a class-balancing weight alpha, be numerically stable for large-magnitude inputs, and support none/mean/sum reduction modes.
  • How would you extend this to multi-class focal loss using softmax over C classes? How does p_t change, and how do you efficiently index the true class probability?
  • Why does the focusing term help even when you already have alpha-balancing? Concretely, how does gamma change the gradient contribution of an easy example with p_t around 0.9 versus a hard one with p_t around 0.1?
  • How do you choose alpha and gamma in practice and how are they coupled? What happens to the loss behavior as gamma approaches infinity?
  • Some implementations stop the gradient through the modulating factor (1 - p_t)^gamma. What is the motivation for doing that, and what trade-off does it introduce?

“My first instinct was to just do sigmoid then log, which is exactly the wrong move.”

View Post

Datadog·Machine Learning Engineer·Technical Phone Screen

Apr 2026
Datadog ML engineer interview with a practical API coding task. They gave you a real integration problem and AI tools were fair game, which was a nice change from the usual whiteboard stuff.
  • Connect to the Snowflake API, submit a query, and retrieve the query status.
  • Follow-up: what would you do if the query returns too much data?

“AI tools were allowed so I leaned on them pretty hard for the auth setup.”

View Post