Sanket Matroja — AI Engineer

Open to Opportunities

that ships.

Full-stack AI engineer. I've built fraud detection engines, recommendation systems, NLP pipelines, and LLM-powered tools — all shipped to production with real architectures, not demos.

View My Work

Resume

14 projects · 5 full-stack platforms · 30+ technologies · Python, TypeScript, React, FastAPI

What I've Built

Full-stack AI systems — from fraud detection to NLP pipelines to recommendation engines.

RiskPulse

AI-powered fraud detection platform with composable rule engine, LLM-generated investigation narratives, and real-time risk scoring across the full Detect-Investigate-Mitigate lifecycle.

▲Next.js ⚛️React 🔷TypeScript 🎨Tailwind CSS 🚀FastAPI

Coffee Sommelier

Algorithmic recommendation engine using weighted cosine similarity and MMR diversification — zero LLM dependency, sub-50ms latency, serving four independent frontends.

▲Next.js ⚛️React 🔷TypeScript 🎨Tailwind CSS ⚡Vite

Mind Mirror

NLP-powered journaling with a 3-tier sentiment pipeline: Hugging Face RoBERTa inference, TextBlob fallback, and LRU-cached analysis with ~40% hit rate.

⚛️React ⚡Vite 🟨JavaScript 🚀FastAPI 🐍Python

ResumeTailor

LLM-powered resume optimization using Anthropic Claude — parses DOCX into structured blocks, intelligently tailors content to job descriptions, and re-exports with formatting preserved.

▲Next.js ⚛️React 🔷TypeScript 🎨Tailwind CSS 🚀FastAPI

SignalVault

GPT-4-powered market intelligence platform aggregating signals from Reddit, Twitter, YouTube, HN, and Google Trends with semantic similarity search.

▲Next.js ⚛️React 🔷TypeScript 🎨Tailwind CSS 🚀FastAPI

MovieMatch

Real-time collaborative filtering for movie matching — Firebase-powered room sync with intelligent preference matching via TMDB data and swipe-based UX.

📱React Native 📲Expo 🔷TypeScript 🔥Firebase 📄Firestore

View all 14 projects

Engineering Stack

Technologies I use daily to build and ship production systems.

How I Think

Architecture decisions, trade-offs, and outcomes behind key projects.

RiskPulse

Designing an AI Detection Engine That Doesn't Need a Data Scientist

Traditional fraud detection systems are rigid — they either rely on static thresholds that generate excessive false positives, or require expensive ML pipelines that are slow to update. RiskPulse needed a detection system that business analysts could configure without engineering involvement, while still supporting complex composite rules.

Rule creation: minutes instead of days5 composable detection strategiesAsync pipeline via Redis workers

Coffee Sommelier

Why I Built a Recommendation Engine Without GPT — And It's Better

Coffee recommendation systems typically either use basic filtering (too simple) or LLM-powered suggestions (expensive, unpredictable latency, non-deterministic). Coffee Sommelier needed recommendations that feel personalized while remaining fast, predictable, and cost-free to operate at scale.

<50ms latency, $0 LLM costCosine similarity + MMR diversificationHaversine geo-filtering

Mind Mirror

Zero-Downtime AI: Building a Sentiment Pipeline That Never Fails

Mind Mirror needs real-time sentiment analysis for every journal entry, but the Hugging Face Inference API has rate limits, occasional downtime, and costs that scale with usage. The system needed to be resilient against API failures while maintaining acceptable sentiment quality.

Zero downtime: 3-tier fallback chain~40% cache hit rate, fewer API calls85% accuracy (RoBERTa) / 70% fallback (TextBlob)

I build AI

What I've Built

RiskPulse

Coffee Sommelier

Mind Mirror

ResumeTailor

SignalVault

MovieMatch

Engineering Stack

Frontend

Backend

Data & Storage

AI / ML

Languages

Infrastructure

Tools

How I Think

Designing an AI Detection Engine That Doesn't Need a Data Scientist

Why I Built a Recommendation Engine Without GPT — And It's Better

Zero-Downtime AI: Building a Sentiment Pipeline That Never Fails