I will evaluate and red team your llm or ai product outputs

E
etzira
E
etzira
Travis Smith
evaluate and red team your llm or ai product outputs
evaluate and red team your llm or ai product outputs
Full Screen
Full Screen
Full Screen
evaluate and red team your llm or ai product outputsevaluate and red team your llm or ai product outputsevaluate and red team your llm or ai product outputsevaluate and red team your llm or ai product outputs

About this gig

I evaluate and red-team AI/LLM outputs for accuracy, safety, and failure modes, using rubric-based scoring, not vibes.


Background: I do this professionally for multiple AI research organizations (under NDA), assessing outputs for instruction-following, factual accuracy, reasoning, and safety, and comparing competing model responses to surface hallucinations and edge-case failures.


Evidence, not adjectives: EndpointDrift, one of my public builds, is a clinical-trial auditor validated against a 200-trial frozen corpus with 358 gold labels and 101 tests. OSINT Fusion runs an adversarial verification swarm across 1,400+ tests. Write-ups at thisistravissmith.com.


What you get: a rubric matched to your product's actual failure modes, a scored pass over your outputs, a written report of specific failures (hallucinations, unsafe completions, reasoning breaks, instruction misses) with examples, and a plain-language severity ranking.


Good fits: pre-launch AI features, chatbot/agent QA, comparing model or prompt versions, guardrail spot-checks.


Send me sample outputs (or sandbox access) and what "good" looks like, and I'll scope the pass.

  • Model expertise

    • Custom model development
    • Generative AI

  • Industry

    • Cyber security
    • Data analytics
    • Financial services
    • Legal

  • Language

    • English

  • Technical expertise

    • Machine learning (Supervised, Unsupervised, Reinforcement)
    • Natural language processing (NLP)
    • AI ethics and bias mitigation

Get to know Travis Smith

Travis Smith

Agentic AI Engineer for LLM Evaluation, RAG, and Custom AI Builds

  • FromUnited States
  • Member sinceJul 2026
  • Avg. response time1 hour

  • Languages

    English
AI engineer building agentic systems for high-stakes, documentation-heavy workflows, with 8 years in clinical documentation before moving into AI. I work AI-augmented: I direct LLM tooling and own the architecture, evaluation, and safety decisions. Current work: LLM evaluation and red-teaming for multiple AI research organizations (under NDA); agentic builds with LangGraph, MCP, and RAG; and custom AI systems for small businesses at Etzira Consulting. Real projects, real test suites, live demos at thisistravissmith.com. Also founder of Exigere Solutions, a medico-legal transcription company.

My Portfolio

Other AI Development Services I Offer

Related tags