GenAI System Design Mock Interview with Shivam Tayal

Interviewing for a GenAI, FDE or Staff role where the loop now includes an LLM system design round — RAG, agents, evals, cost? This is a live mock of that round, run by someone who has built and shipped these systems.

Only for engineers interviewing for senior, Staff, FDE or AI engineer roles — this isn't an intro to LLMs.

In this session:

• 45 min — you design a production GenAI system: RAG search over enterprise docs, a tool-using agent workflow, or an LLM feature under a strict latency and cost budget. I push on retrieval quality, evals, guardrails, failure handling, cost at scale, and where an LLM shouldn't be in the path at all.

• 15 min — debrief with a scored rubric (problem framing, context and retrieval design, evaluation, reliability, cost) and the 3 gaps to fix before your loop.

Why me: At Coinbase I led the team that built a LangGraph multi-agent platform processing 10M+ GitHub events/day across 850+ repos — that system cut LLM inference costs 70% (~$680K/year), with testing built on golden-dataset regression and LLM-as-judge evaluation. Before that, 5 years at Media.net building high-throughput ad-fraud systems (25M+ ad views/day, Kafka + Spark). This is production experience, not a weekend project.

Book below — bring the GenAI system you'd be asked to design.