Open to opportunities

Harshad Nagpure

@harshadnagpure1

Message

I build reliable AI, computer vision, and networking systems from complex real-world data.

India

Message

I built a DNS server in Go with RFC 1035-compliant query resolution, packet parsing, header compression, caching, and concurrent request handling.

I've also turned unstructured credit-card statements into structured JSON through a resilient Python extraction pipeline and Streamlit analytics dashboard, including spending analytics and heuristic fraud detection.

For hearing-impaired users, I built a real-time Indian Sign Language-to-text converter using 7,000+ gesture images and 42 hand landmarks, reaching 99.8% accuracy while reducing inference time by 40%. I'm pursuing a B.Tech in Information Technology and bring foundations in machine learning, computer vision, networks, and data systems.

Experience

Work history, roles, and key accomplishments

Credit Card Statement Analyzer

Personal Project

Engineered a zero-fail extraction pipeline across five major issuers using cascading logic: PyMuPDF, OCR, and Gemini LLM fallback. Transformed unstructured financial PDFs into structured JSON for analytics and implemented smart spending analytics and heuristic fraud detection in a Streamlit dashboard.

Python Streamlit PyMuPDF Gemini API Pandas

DNS Server Developer

Personal Project

Developed a fully functional DNS server in Go implementing packet parsing, header compression, and RFC 1035-compliant query resolution. Engineered core modules for message encoding/decoding, caching proxy, and concurrent query handling using goroutines and channels.

UDP TCP Concurrency Networking Protocols

Real-Time Sign Language Converter

Personal Project

Built a real-time ISL-to-text system using 7,000+ gesture images and 42 hand landmarks, achieving 99.8% accuracy. Optimized a low-latency OpenCV pipeline, reducing inference time by 40%, and added multilingual translation for hearing-impaired users.

Python OpenCV Mediapipe scikit learn