At Mercor, I write domain-specific evaluation scenarios and gold-standard SFT references, and red-team frontier models through multi-turn adversarial conversations in English and Spanish. I document reproducible exploit steps, severity ratings, and reasoning flaws across AI safety and evaluation contracts.
At MXS Solutions, I tested staging environments and documented reproducible bugs, helping cut recurring engineering escalations by 35%. Earlier, at Fiserv, I used SQL and Postman to validate transaction data and API payloads across financial service releases.

