At Solix Technologies, I built Prompt2SQL, an LLM-powered Text-to-SQL copilot that generates dialect-aware SQL across Oracle Database, PostgreSQL, Apache Hive, Databricks, and the IBM Db2 family. I grounded generation on Oracle E-Business Suite and SAP schemas using vector databases, hybrid retrieval, and LLM re-ranking, then served it as a FastAPI microservice.
I also engineered Virtual Screening, a drug-discovery ML pipeline that predicts compound-target binding affinity from SMILES. I used RDKit and Mordred to generate molecular descriptors and fingerprints, then applied feature selection before scikit-learn model training and evaluation.
Earlier at Solix Technologies, I built RAG prototypes for enterprise search and logging and monitoring for distributed AI services. My projects include a configurable RAG framework and an IDP benchmark of OCR and parsing libraries, as well as embedded systems projects in assistive communication and robotics.
