Skip to content

About Me

I'm Sanket Matroja. I build AI-powered products that actually work in production. Not prototypes. Not demos. Shipped systems used by real users.

My work spans 14 projects using 31+ technologies — from AI-powered fraud detection engines (RiskPulse) to algorithmic recommendation systems with zero LLM cost (Coffee Sommelier), to NLP sentiment pipelines with 3-tier fallback chains (Mind Mirror), to GPT-4 signal intelligence platforms (SignalVault).

AI Philosophy

AI is a tool, not a feature

I don't add AI for the buzzword. Every model, every pipeline, every LLM call solves a specific user problem.

Deterministic when possible, intelligent when necessary

I built a recommendation engine with zero LLM cost using cosine similarity. I use GPT-4 when the problem demands it. I always know why.

Resilient AI pipelines

My systems don't break when the API goes down. 3-tier fallbacks, LRU caches, graceful degradation. Zero downtime for AI features.

Ship, then optimize

I get AI into production fast, measure what matters, and iterate. No research rabbit holes. No perfect-model paralysis.

Core Stack

AI/MLOpenAI GPT-4, Anthropic Claude, Hugging Face, TextBlob, Sentence Transformers
FrontendReact, Next.js, TypeScript, Tailwind CSS, React Native
BackendFastAPI, Python, Node.js, Celery
DataPostgreSQL, MongoDB, Redis, Supabase, Firebase
InfrastructureDocker, AWS Lambda, Vercel