At Mercor, I train and evaluate AI models on Indic-language scripts, checking outputs for accuracy, consistency, and edge cases. I identify model failures and provide structured feedback to improve training workflows and output quality.
I built Teams Accountability Agent, an end-to-end LLM application that turns meeting transcripts into action items, decisions, and accountability insights. I improved its reliability with structured outputs, schema validation, fallback parsing, and human review.
For Unsupervised Anomaly Detection for Structural Health Monitoring, I developed an ML pipeline for vibration data across 17 structural scenarios and 33 sensor channels. I improved signal processing and compared detection methods, achieving 0.98 AUROC and 95% recall.
I also built Quant Quest, a quantitative finance platform with seven games, virtual wallet transactions, and real-time leaderboards for 400+ participants. I’m a founding member and Networking/Marketing Head of Quantix at IIT Madras, and a Top 4 Finalist in the Fetch.ai AI Agent Hackathon.
