At Frost Digital Ventures, I own the build, debugging, and QA of a multi-tenant cybersecurity threat-intelligence pipeline. It combines NVD, MSRC, and CISA KEV data with Qdrant hybrid retrieval and generates tenant-specific risk reports using OpenAI GPT models.
I'm also building webhook-driven FastAPI services for WhatsApp customer support, AI-assisted email replies, and voice interactions, with human escalation where needed. I built an invoice-extraction service for photographed and scanned Nepali business documents, with OpenAI vision and local Ollama backends.
Before Frost Digital Ventures, I built and maintained R and SAS ETL pipelines on anonymized clinical-trial datasets at Numeric Mind, following CDISC standards. I also performed statistical modeling and exploratory analysis on datasets of 10,000+ records.
My computational mathematics background informs projects including a breast cancer prediction model that reached 97% accuracy and a comparison of dimensionality-reduction methods. I'm interested in building practical AI systems grounded in reliable data and thoughtful human review.
