At Farmitrx, I lead AI/ML architecture and engineering across AgroPro AI and Secure Agentz AI, translating product and domain requirements into integrated systems. For AgroPro AI, I architected workflows spanning agronomic records, recommendations, compliance review, document generation, and regulatory filing.
I integrated local Parakeet-TDT speech recognition and Qwen3 intent routing with structured entity resolution and deterministic application logic. The platform uses a multi-tenant PostgreSQL model and FastAPI backend, with immutable document snapshots that preserve reproducibility after a recommendation is signed.
For Secure Agentz AI, I designed the architecture and technical direction for a local-first multi-agent platform with defined control, execution, and learning responsibilities. I also developed multimodal retrieval for complex document collections, combining structure reconstruction, embeddings, Milvus retrieval, and filename/page provenance for grounded generation.
