Brijesh Kumar

A scalable AI-powered educational platform engineered with Next.js 14 and FastAPI, integrating Azure OpenAI and Groq to deliver 92% accurate personalized content in 3.2s. Designed real-time analytics and telemetry pipelines processing 50K+ events, improving engagement by 85% and boosting retention by 78%.

A full-stack career platform utilizing RAG and vector embeddings to deliver personalized resume-job matching, increasing application rates by 40%. Engineered resilient microservices for auth and search handling 180 RPS with zero errors, and implemented automated Java web scraping to expand job databases by 3x.

A high-performance recommendation engine using CLIP and ResNet50 for vector-search, boosting user engagement by 32%. Features a real-time React+TypeScript swipe UI backed by an optimized FastAPI service with ANN, achieving sub-150ms p95 latency for 10k+ users with 99.99% uptime.

An intelligent contact management system leveraging Spring Boot and Kafka for microservices orchestration. Integrated AWS Bedrock and FAISS for RAG-based contact enrichment, raising match accuracy by 70%. Optimized with Redis read-through caching to deliver sub-3s network summaries.

An intelligent computer vision platform leveraging deep learning and real-time image processing to analyze visual data with high precision. Architected and deployed a scalable full-stack system integrating AI inference pipelines with an optimized Next.js frontend for seamless user interaction. Engineered modular, production-ready APIs to enable low-latency predictions, robust cross-device performance, and future model extensibility. Designed with scalability, reliability, and deployment efficiency at its core.

An AI-powered learning assistant that transforms dense concepts into short, dialogue-based explainer videos with quizzes, streaks, and badges. If a student doesn’t understand a concept, SlideSage instantly regenerates fresh analogies until it clicks — making learning adaptive, fun, and addictive.

A hybrid RAG healthcare system combining semantic and keyword search with Sentence-BERT, improving top-k precision by 75%. Deployed a local LLaMA2 model with strict context-only guardrails to achieve 100% test-query accuracy, served via a robust Flask API on AWS ECS.

A production-grade multi-agent Retrieval-Augmented Generation (RAG) system designed for high-precision medical query handling. Engineered an adaptive two-threshold orchestration algorithm improving retrieval precision by 40% over single-agent baselines. Implemented parallel agent optimization using ThreadPoolExecutor and utility-based scoring, reducing query latency from 12s to 4s (3x speedup). Integrated Sentence embeddings with ChromaDB vector search and LLM-powered metadata scoring achieving 0.89 similarity precision. Deployed scalable FastAPI and ReactJS services with OpenAI integration on AWS, maintaining 99% uptime.

An end-to-end data pipeline ingesting tracks via Spotify API, automating ETL to reduce manual prep by 80%. Utilized Tableau and Python for EDA to identify viral trends and trained Scikit-learn models predicting playlist inclusion with 0.92 R2 score.