Senior AI Engineer
Posted Aug 5, 2026 · We last checked this listing on Oct 5, 2026
Likely interview questions for this role
Written from this job description, not a generic list. Each one notes what the interviewer is really checking.
About your past
Tell me about the most significant ML system you took from prototype to production.
What they are checking: Real end-to-end ownership, including the unglamorous operational work after launch.
The AI function is new and central. How would you work with teams who already use frontier APIs and do not want to switch?
What they are checking: Whether they earn adoption by proving value instead of forcing a migration.
What would you do
DV wants to own its model capability instead of depending on frontier providers. How would you decide which tasks are worth moving to fine-tuned open-weight models first?
What they are checking: Whether they can pick early wins by weighing volume, cost, latency and how well a smaller model can do the task.
A fine-tuned model starts giving worse answers a month after launch. How do you investigate?
What they are checking: Whether they look at input drift, data changes and monitoring rather than retraining blindly.
How would you explain to leadership the cost trade-off between buying GPUs and paying per token?
What they are checking: Ability to model total cost of ownership and communicate it in business terms.
About the work
Walk me through how you would build a model gateway that routes inference across open and closed models by cost, latency and quality.
What they are checking: Design depth on routing rules, fallbacks, observability and a common interface that hides provider differences.
How would you fine-tune and then distill an open-weight model for a DV-specific task?
What they are checking: Hands-on knowledge of data preparation, methods like LoRA, distillation and how to judge the result.
What does it take to run inference on-prem with good throughput and latency?
What they are checking: Familiarity with serving stacks like vLLM, batching, quantization and GPU memory limits.
How do you prove a distilled model is good enough to replace a frontier model in production?
What they are checking: Use of task-specific evaluation sets and live comparisons rather than general benchmarks.
How would you handle confidential trading data when training or serving models?
What they are checking: Awareness of data access, logging and leakage risks in a proprietary trading firm.
Practice this interview out loud.
Offer builds a real interview for this exact role at DV Trading from your resume and this job description, asks the questions one at a time, and tells you what landed. The first one is free.
Practice this out loudThe full job description
As published by DV Trading.