At Hughes Systique Corporation, I designed and built a LangGraph Q&A chatbot over an 11K-page technical specification corpus. It achieved 93% end-to-end answer accuracy across 500 TSPEC test queries.
I also built an enterprise RAG pipeline over technical documents, emails, and meeting transcripts, improving Recall@K to 85% and Precision@K to 83%.
As an Associate AI Engineer at Hughes Systique Corporation, I redesigned retrieval for technical documentation and reduced latency by 10× while maintaining a Recall@10 of 0.88. I also developed a PyTorch architecture for time-series signal classification that achieved 87% accuracy on 1,000 real signal samples.
I fine-tuned FLAN-T5 for domain-specific question answering with QLoRA/PEFT, reducing GPU training cost by 60%, and deployed it through a real-time inference API. In a project, I built a LangGraph-powered email automation platform with human-in-the-loop approval gates.
