Highlights

• Built zero-to-one multi-agent products processing 1,000+ hazardous waste documents, generating $500k+ ARR with 20% monthly growth
• Led end-to-end ML product development with CI/CD, A/B testing, achieving 95%+ accuracy on document classification
• Fine-tuned LLMs for waste classification, improving recall and precision 50%+ vs third-party foundation models
• Designed knowledge graph systems integrating federal/state regulatory databases, reducing legal review time by 70%