Practical insights on machine learning, LLMs, causal inference, and building AI systems in production.
Point your phone at the TV and ask anything. How I built an AI companion that identifies shows from camera captures and answers contextual questions — using Gemini vision, TMDB, and voice input.
The story behind ArguLens — from a friend's divorce paperwork to an AI product that finds contradictions, reconstructs timelines, and grounds every claim in source documents.
TalentGrep analyzes GitHub profiles to surface real skill profiles — moving past resume keywords to what engineers have actually shipped.
How we built an AI system that monitors 150K+ live sales calls per month, catching compliance violations that keyword-based systems miss entirely.
How we moved past "who will churn" to "what will prevent churn" — using causal graphs, propensity matching, and targeted A/B tests to drive a 160 basis point reduction worth $1M annually.
EEG signal processing, swarm robotics, medical imaging, then data science. Every detour added a tool I use daily — and non-linear careers produce the best engineers.
Building a real-time propensity model that matches incoming customers to the right sales agent — a 12% conversion lift that changed how our entire call center operates.
Using DSFD, InsightFace, and DBSCAN to surface the photos people actually care about — and how it moved 30-day retention by up to 150 basis points.