Lead AI Systems Trainer & Workflow Architect — Valiant Inc.
Trained and aligned open-source and proprietary LLMs by generating, labeling, and curating structured textual RLHF datasets. Used prompt engineering to reduce hallucinations and improve answer quality for B2B applications. Performed evaluation-oriented checks using adversarial testing to verify logical correctness, safety, and factual consistency. • Created labeled RLHF textual datasets aligned to target linguistic benchmarks • Designed multi-shot, chain-of-thought, and role-based prompt/system instruction frameworks • Ran adversarial (red-teaming) fact-checking to expose edge cases and correct outputs • Verified accuracy against linguistic, logical-flow, safety, and brand-alignment criteria