At Western EcoSystems Technology Inc, I productionized more than 10 multimodal computer vision systems and deployed a OneFormer model on an NVIDIA Jetson Nano for real-time wildlife segmentation. I also engineered an embedded ML workflow for carcass detection from infrared drone imagery, reducing field-survey costs by 60%.
I deployed a Llama2-based RAG pipeline to summarize technical RFPs, eliminating more than 400 hours of manual analysis. I also reduced annotation effort by 45% and accelerated model deployment speed by 2x through improvements to pre-labeling and CI/CD workflows.
At Stevens Institute of Technology's AISec Lab, I built a data pipeline for suicide-risk sentiment detection and analyzed cascading failures in transmission outages. Earlier, at L&T Technology Services and Persistent Systems, I worked on backend systems, APIs, databases, and production monitoring.
