Consumer harm intelligence
HarmScope
NLP pipeline over 16.5M CFP complaints and 3.8M narratives to identify emerging consumer harm patterns.
02 / Selected work
A fuller record of tools, experiments, and research systems built across campus operations, applied AI, data analysis, and language.
Consumer harm intelligence
NLP pipeline over 16.5M CFP complaints and 3.8M narratives to identify emerging consumer harm patterns.
Campus product
Launched a real-time NFC dining platform supporting 1,000+ daily NFC check-ins, helping students locate available seating across campus.
Applied AI
Processed 500K+ FDA adverse-event reports using NLP and clustering to identify high-risk devices and accelerate safety analysis.
Language systems
Created an open-source editorial lexicon documenting Sanskrit-derived Hindi vocabulary through searchable linguistic relationships.
Research
Trained computer vision models on 2,000+ UAV images, improving wildfire detection accuracy for low-visibility environments.
Language research
Open research lab building a versioned Garhwali corpus, benchmark, tokenizer studies, and language models for low-resource language research.
Embedded safety
Built an Arduino smart helmet that detects accidents and sends live GPS coordinates by SMS, with false-positive mitigation, a cancel window and multi-contact alerts.
Financial data analysis
Analyzed 72,000 congressional stock trades across 350+ members using OCR, fuzzy matching, and statistical testing.
Applied AI
Analyzed 50 B2B survey transcripts with NLP, sentiment scoring and K-Means clustering to build five buyer personas and actionable go-to-market recommendations.