Gen AI Lead

Pleasanton, CA

Resume/CV *

Attach

First name

Last name

Phone

Can you work onsite in Pleasanton, CA on a hybrid schedule (3 days per week in office)?

Yes

Are you available to work on a contract basis?

Yes

Walk us through a production Gen AI system you personally architected and shipped, from the raw business problem to live traffic. Be specific: what model(s) you selected and why, how you handled the RAG vs. fine-tuning vs. prompt-engineering trade-off, what your token economics looked like at scale, and your p95 latency and cost-per-request in production. Then tell us the one architectural decision you got wrong and what it cost the business before you caught it.

Describe the evaluation framework you built to measure and control hallucination in a real deployment — not a benchmark you read about. How did you define "correct," what did your golden dataset and human-in-the-loop process look like, what guardrails or grounding techniques did you implement, and what was your measured hallucination rate before and after your interventions? If you cannot give numbers, explain honestly why not.

You inherit a Gen AI application that costs $180K/month in inference, has 40% user abandonment from latency, and leadership will kill it in 60 days. Walk us through exactly what you do in week one, what you measure, which levers you pull first (distillation, caching, routing, quantization, batching, or other), and how you reach a defensible save-or-sunset decision. Be concrete about sequence and trade-offs — "it depends" is not an answer.

Describe how you have handled PII, prompt injection, data leakage, and model access controls in a regulated or security-conscious environment. What was your position on third-party model APIs vs. self-hosting, and how did you win — or lose — that argument with security and legal? Give us the real friction, not the sanitized version.

Tell us about a time you were the most senior Gen AI person in the room and had to bring along skeptical engineers or executives who either overestimated the technology (thought it was magic) or underestimated it (thought it was a toy). How did you calibrate expectations, and what evidence did you use to move them? If the project underdelivered against the hype, own your part in setting or failing to reset those expectations.

We use cookies to offer you our service. By continuing to use this site, you consent to our use of cookies as described in our policy