I am Krish Ujeniya — an AI Researcher, Data Scientist, and independent AI Freelancer based in India. I help teams design, evaluate, and deploy production-grade machine learning models, autonomous agent workflows, and reproducible data pipelines.
My engineering work emphasizes practical reliability: training high-accuracy models in PyTorch and Scikit-Learn, implementing strict data lineage with DVC, building sub-150ms vector search systems, and shipping secure, well-tested Python microservices.
|
Working production implementations across multi-agent workflows, healthcare speech recognition, computer vision, and machine learning pipelines.
🚀 Hadron ↗Autonomous Multi-Agent PR Review Engine. Automated code review system for GitHub pull requests. It extracts abstract syntax trees (ASTs), runs hybrid DiskANN vector search over the repository dependency graph, and runs specialized verification agents to flag race conditions, security vulnerabilities, and logic flaws before merge.
⚡ Live Demo ↗ • 📁 GitHub Repository ↗ • 📄 Technical Report ↗ |
|
Developer Monorepo for Autonomous AI Agents. Collection of developer tools and deterministic frameworks for autonomous agents. Includes the Anchor 5-gate verification pipeline, a multi-agent deliberation CLI, and standard Model Context Protocol (MCP) server integrations for local tool execution.
📚 Documentation Site ↗ • 📁 GitHub Repository ↗ • 🎓 Citation CFF ↗ |
|
Lightweight Personal AI Assistant Framework. Connects LLMs to local MCP tools with persistent conversational memory across Telegram, Discord, and Slack in under 4,000 lines of readable Python. Features SQLite session storage and asynchronous tool execution.
📁 GitHub Repository ↗ • 📊 Benchmark Suite ↗ • 🏛️ Architecture ↗ |
|
Voice-to-Form Clinical Visit Entry System. Transcribes medical conversations into structured clinical encounter records using Gnani Prisma and Timbre ASR models. Prompts healthcare staff with targeted clarification questions whenever drug names, dosages, or instructions are ambiguous.
|
🛡️ WaferGuard ↗Semiconductor Wafer Anomaly Detection & MLOps Pipeline. End-to-end machine learning system for semiconductor defect classification. Implements automated validation gates, reproducible pipeline stages with DVC, and metric tracking across model iterations to catch wafer defects early.
📁 GitHub Repository ↗ • 🔄 DVC Lifecycle ↗ • 🎓 Citation CFF ↗ |
👁️ Hajri.ai ↗Computer Vision Attendance System with Anti-Spoofing. Real-time facial recognition service with active liveness verification. Analyzes facial micro-movements to reject printed photo attacks and screen replays in institutional attendance workflows.
|
|
On-Device Mobile OCR Application. Cross-platform Flutter app that extracts structured text from camera photos and saved images. Runs 100% locally using Google ML Kit on Android and iOS with zero cloud network calls and instant text sharing.
|
|
Automated Network Reconnaissance Tool. High-speed Bash orchestration script wrapping Nmap for network host discovery, port audits, service versioning, and operating system detection. Exports cleanly formatted, timestamped logs for penetration testing audits.
|
|
LangGraph Cyclic State Graphs · Model Context Protocol (MCP) · LangChain · DiskANN Reciprocal Rank Fusion (RRF) · Multi-Agent Deliberation CLI · Retrieval-Augmented Generation (RAG) · LiteLLM Unified Routing · Prompt Engineering & Context Economics |
|
PyTorch · Scikit-Learn · DVC (Data Version Control) Pipeline Tracking · SHAP Interpretability & Feature Attribution · OpenCV Computer Vision · Biosignal Telemetry · Sensor Fusion Human Activity Recognition (HAR) · Time-Series Regression |
|
Python 3.12 · FastAPI Asynchronous Web Framework · PostgreSQL Relational Databases · SQLite Embedded Storage · Redis In-Memory Cache · Docker Containerization · Bash Shell Automation · Linux Microservices · GitHub Actions Automated CI/CD |
|
Flutter Cross-Platform Development · Dart · Next.js 15 App Router · React 19 · TypeScript · Tailwind CSS · On-Device Google ML Kit · Responsive Glassmorphism Mobile-First Design |
|


