Anuj Umanath Jha
AI/ML engineer and system builder focused on rapid prototyping of agentic, real-time AI solutions. Strong experience delivering end-to-end products in hackathons using LLMs, RAG, and predictive modeling.
Education
SVKM’s DJ Sanghvi College of Engineering
TE - B. Tech
Skills
PythonCJavaSQLPostgreRedisMongoDBChromaDBpgvectorAgentic AIRAG PipelinesLang ChainGenerative AIScikit-learnOpenCVPandasNumpyTime-Series ForecastingHTMLCSSTailwind CSSGitDocker
PROJECTS
Autonomous Multi-Agent Research Analyst
Built a cyclic multi-agent research system using Lang Graph with specialized search, synthesis, and fact-check agents that auto-verify claims against raw sources and loop back if confidence < 80% Enforced structured citations via Pydantic schemas with >90% verification rate; deployed real-time Agent Trace dashboard on Hugging Face Spaces
Real-Time Neural Hand Controller
Trained a sub-100 KB CNN via knowledge distillation from Mobile NetV 3 for 5-class gesture recognition, achieving 94% accuracy at <16 ms latency in browser via Tensor Flow Built live Grad-CAM interpretability overlay exposing pixel-level model reasoning; deployed interactive gesture-control interface with confidence thresholding as static site
Arogya – AI Hospital Operations Platform
Built a multi-agent AI system for hospital operations, decision support, and alerts using agentic workflows and RAG. Delivered a real-time full-stack platform with multilingual notifications and voice assistance for accessibility.
Cognito – AI-Driven Gamified Learning Platform
Developed an AI-powered adaptive learning system that generates non-repeating quiz questions and adjusts difficulty in real time based on learner performance. Integrated instant AI-generated feedback with competitive gamification to improve engagement and knowledge retention.
SENTINEL – AI-Based Satellite Collision Risk Prediction
Developed an AI-driven system to analyze orbital trajectories and predict potential satellite collision risks using probabilistic reasoning. Generated proactive risk alerts through real-time inference on dynamic orbital data streams.