ML Engineering interview question · ML system design · staff level
An online model metric dropped 8% overnight. How do you debug it?
Free to read in full. This is 1 of the 150 questions in the ML Engineering and MLOps interview pack, with 193 diagrams. Get the pack, ₹999 · $20
Why it is asked
Production incidents test judgment under pressure. The interviewer looks for an ordered search that rules out cheap explanations first and restores service quickly.
The answer
Treat it as an incident: confirm the drop is real, find what changed, narrow it to a segment, check the inputs, and only then suspect the model. A sudden overnight drop usually has a sudden cause, such as a release or a broken pipeline, while model decay tends to be gradual. 1. Is the metric right? Check the logging and dashboard pipeline: late or missing event data, a changed metric definition, a tracking bug in a new app version, missing or duplicated partitions. Compare with an independent source, such as orders in the transactions database. 2. What changed in the last 24 hours? List every deploy: model version, serving code, feature pipeline, config, thresholds, new experiments, and upstream services. If the drop starts exactly at a deploy, roll back first and investigate after. 3. Where is the drop? Slice by platform, app version, country, traffic source, new versus returning users, and fallback rate. An 8% drop that sits entirely in one app version is a client bug; a drop spread evenly points to data or model inputs. 4. Are the inputs healthy? Compare today's feature distributions and null rates with last week's; check feature freshness, and whether online values match offline values for the same entity (training-serving skew). A stale or empty feature is a frequent cause. 5. Did the world change? Holidays, a marketing campaign bringing new kinds of users, an outage at a partner. Compare with the same day last week. 6. Then the model: score yesterday's traffic with the previous and current model to compare. Throughout: keep a written timeline and, afterward, add an automatic check so the same cause alerts next time.
What the interviewer listens for
- Verify the metric pipeline before touching the model.
- Correlate with deploys; roll back first if one lines up.
- Slice to localize the drop.
- Check feature health, freshness and skew.
- External causes, then the model; add a check afterward.
Follow-ups the interviewer may ask
- How do you know if it is seasonal? Hint: compare with the same weekday last week and the same period last year, and with a holdout group.
- No deploy, no segment, inputs look fine: what next? Hint: look at upstream product changes and traffic mix, then compare model versions offline on yesterday's data.
- How do you prevent a repeat? Hint: automatic checks on null rates, freshness and per-segment metrics, with alerts tied to deploys.
Common mistakes
- Retraining immediately without knowing the cause.
- Debugging the model for hours while a recent deploy sits un-rolled-back.
- Trusting the dashboard without checking the logging pipeline.
Written by Roni Das, who builds distributed systems and AI systems for a living.
Learn it properly
This is taught step by step in the course lesson Failing Silently: Which Checks Catch a Broken Input Before the Answers Arrive.
One question is not enough for the interview.
This is 1 of the 150 questions in the ML Engineering and MLOps pack
Each question has why it is asked, a full answer, what the interviewer listens for, the follow-ups, the common mistakes and a diagram (193 diagrams in all). A PDF you keep, junior to staff level.
India: UPI, cards or netbanking through Razorpay. Elsewhere: PayPal. Download at once from My packs.
More free questions from this pack
- A model scores 95% accuracy offline but users say it is useless. What do you check first? (mid)
- How do you train on a dataset that does not fit in memory? (junior)
- A long distributed training job dies at hour 20. How do you make training fault tolerant? (senior)
- How does request batching on a GPU trade latency for throughput? (mid)
Free questions from the other packs
- What do temperature and top-p control, and what values would you use for data extraction versus brainstorming? (AI Engineering)
- What is hybrid search, and how do you merge keyword and vector results with Reciprocal Rank Fusion? (AI Engineering)
- Users ask "list all X" questions and RAG returns only five items. Why, and what do you do? (AI Engineering)
- How do you check that an LLM judge agrees with human reviewers? (AI Engineering)
- Cut the cost of an LLM feature serving 10 million requests a day by 60% without hurting quality. Walk through your plan. (AI Engineering)
- Why are LLMs bad at counting letters or reversing a word? (GenAI and LLM)
- Write out scaled dot-product attention with the shape of every tensor. (GenAI and LLM)
- How does speculative sampling keep exactly the target model's distribution? (GenAI and LLM)
- What is GRPO, and how do verifiable rewards train reasoning models? (GenAI and LLM)
- Design a system that answers questions from 300-page contracts accurately. (GenAI and LLM)
- Describe the basic agent loop. (Agentic AI)
- What are pass@k and pass^k, and why do they matter for agents? (Agentic AI)
- How does prompt injection become dangerous once the agent has tools? (Agentic AI)
- Design the security architecture for an agent that can read and send company email. (Agentic AI)
- Design a customer support agent for an e-commerce company. (Agentic AI)
- What happens, step by step, when an LLM server handles one request? (Inference Engineering)
- Derive the KV cache size formula and work it out for Llama 3 70B. (Inference Engineering)
- What is PagedAttention? (Inference Engineering)
- When does speculative decoding stop helping or make things worse? (Inference Engineering)
- Design a service to serve Llama 3 70B at 10 QPS with latency targets. (Inference Engineering)
- You need to show delivery times for 20 cities. Would you use histograms, box plots, violin plots or ECDFs, and why? (Data Science)
- A test is 99% accurate and you test positive. What is the chance you have the disease? (Data Science)
- What share of first-time buyers make a second purchase within 30 days? Write the SQL. (Data Science)
- How would you forecast daily orders for the next four weeks, and how do you know the forecast is any good? (Data Science)
- A coupon A/B test shows a small average lift. How do you find which users the coupon actually works for, and who should get it next time? (Data Science)
- Compute the cosine similarity of a = (3, 4) and b = (4, 3) by hand. What does the result mean? (Maths for ML)
- A batch of 32 MNIST images goes into a dense layer with 128 units. What are the shapes, and what is the rule for when two matrices can be multiplied? (Maths for ML)
- A disease affects 1% of people. A test catches 90% of sick people (sensitivity) and wrongly flags 9% of healthy people (false positive rate). You test positive. What is the chance you are sick? (Maths for ML)
- Why do we not use mean squared error for classification with a sigmoid output? Compare the gradients numerically. (Maths for ML)
- Compute scaled dot-product attention by hand for two tokens with d_k = 2. (Maths for ML)
- A customer says: "We want AI for our support team." What do you do in the first meeting? (Forward Deployed Engineer)
- The customer's admin offers you a global admin account "to save time". What do you ask for instead, and why? (Forward Deployed Engineer)
- Your connector writes tickets into the customer's system. A request timed out, the connector retried, and now there are duplicate tickets. How do you make the writes safe to retry? (Forward Deployed Engineer)
- The customer wants RAG over 10 million documents, all behind their firewall. Size it for me on the whiteboard. (Forward Deployed Engineer)
- One deployment of your AI platform must serve 50 business units of the same enterprise. How do you design multi-tenancy? (Forward Deployed Engineer)