Shubham Pawar
Full-Stack AI Engineer
Shubham is a Full-Stack AI Engineer with practical expertise spanning computer vision, deep learning pipelines, RAG retrieval platforms, and production web applications. He builds end-to-end intelligent systems, combining custom vision and model integration with high-performance async backends and interactive frontend interfaces.
With a strong computer engineering background and published research in Springer journals on intelligent systems, Shubham focuses on bridging cutting-edge AI models—from OpenAI Vision and custom YOLO models to LLM orchestration—into scalable, enterprise-ready software.
What Shubham works on
Computer Vision & Edge AI
Builds real-time vehicle detection, object tracking, and vision pipelines using YOLOv8, MediaPipe, and OpenCV for production deployment.
RAG & LLM Platforms
Engineers semantic code generation and document intelligence platforms using FAISS, vector embeddings, document chunking, and LLM synthesis.
Full-Stack AI Applications
Develops reactive web apps with Next.js/React and FastAPI backends, integrating real-time WebSockets, SSE streaming, and async task queues.
Document & Vision Intelligence
Constructs production OCR pipelines using OpenAI Vision, PaddleOCR, and custom validation for structured data extraction.
Tools and technologies
AI & Machine Learning
- Computer Vision
- Deep Learning
- YOLOv8 & MediaPipe
- PyTorch & TensorFlow
- RAG Systems
- LLMs & LangChain
- OpenAI & Gemini APIs
- OpenCV
Backend & Infrastructure
- Python
- FastAPI
- Flask
- PostgreSQL
- FAISS Vector DB
- Redis
- Docker
- Cloudflare
- AWS
Frontend
- React
- Next.js
- Remix
- TypeScript
- JavaScript
- React Native
- HTML/CSS
- Tailwind CSS
DevOps & Systems
- Kubernetes
- CI/CD Pipelines
- WebSockets & SSE
- OAuth2 & JWT
- MicroPython
- Raspberry Pi & ESP32
Selected work
- 01
Full-Stack production AI web applications
Architected and shipped AI-powered web applications using Python, FastAPI, Next.js, PostgreSQL, Docker, AWS, and OpenAI/Gemini APIs across multiple production software products.
- 02
RAG-based AI code synthesis platform
Engineered FAISS vector embedding retrieval, document chunking, semantic search, and prompt orchestration for LLM-powered code generation.
- 03
Real-time computer vision & vehicle speed estimation
Built computer vision pipelines using YOLOv8, MediaPipe, and OpenCV for real-time vehicle detection, centroid tracking algorithms, and speed estimation.
- 04
Conversational voice AI calling system
Designed an end-to-end voice calling pipeline featuring RAG-based context retrieval, speech-to-text (STT), LLM dialogue management, text-to-speech (TTS), and real-time call orchestration.
- 05
Production OCR & document intelligence pipelines
Created multi-model vision OCR using OpenAI Vision and PaddleOCR with custom validation for automated structured data extraction.
- 06
Dynamic Traffic Control System (DTCS)
Designed an AI system for real-time vehicle recognition from live camera feeds integrated with IoT devices; published in Springer journal, won 1st place at project expo, and received ₹90K state SSIP funding.
- 07
SkyChords - AI music generator
Developed an AI-driven melody composition algorithm with 90% accuracy, MIDI file generation, and audio streaming using PyTorch and React Native.
- 08
Asynchronous streaming AI microservices
Engineered high-concurrency AI services using background workers, WebSockets, Server-Sent Events (SSE), and Redis caching for real-time user experiences.
- 09
Phishy - ML phishing website detection
Built a phishing detection model achieving 92% accuracy using feature extraction and Random Forest/Decision Tree classifiers.
