I'm Shubham Kulkarni, building production-grade Computer Vision & Generative AI systems at Arya Omnitalk — powering live highway infrastructure processing 50,000+ events daily with 95% model accuracy.
I'm Shubham Kulkarni, an AI Engineer at Arya Omnitalk Wireless Solutions in Pune. I design & deploy production Computer Vision & Generative AI pipelines powering India's live highway monitoring — VIDES & ATMS systems.
My systems handle 50,000+ events daily at 95% accuracy with inference under 100ms. I've engineered 30× faster vector retrieval via FAISS IVF indexing, serving 150,000+ forensic vehicle queries/day.
National Champion (AIR 1) at Smart India Hackathon 2023 as Team Lead. B.E. in Electronics & Telecommunication from SPPU, 2025.
From internships at IBM & MathWorks to building production AI at scale.
Building VIDES + ATMS highway monitoring systems. 50K+ events/day at 95% accuracy. Optimized inference to <100ms. Vehicle forensic matching handling 150K+ daily queries.
Applied LLMs for agriculture & science analytics. Built data pipelines for large-scale insight generation.
Organized hackathons, coding contests, and workshops. Built developer community on campus.
Built NLP text classifiers achieving 90% accuracy using transformer architectures on production datasets.
Developed deep learning & image processing solutions using MATLAB. Computer vision algorithm design.
Alumni Portal deployed live. Core Committee Secretary. Led team to AIR 1 at SIH 2023.
End-to-end AI — from training models to deploying on edge devices at production scale.
Production AI systems, hackathon winners & open-source tools.
Production-ready Document AI + RAG platform with hybrid retrieval (FAISS + BM25 + RRF), streaming Q&A, knowledge graph explorer, and analytics dashboard. Supports PDF, DOCX, XLSX, images & web URLs.
Production-grade vehicle detection & tracking for AI training datasets. Multi-camera RTSP streams, ByteTrack + Kalman filtering for zero-duplicate tracking, ROI-based filtering, and quality control.
IoT-powered irrigation with AI-driven water optimization. Uses real-time soil moisture sensors + weather APIs processed by ML model. AIR 1 at Smart India Hackathon 2023.
Full-stack role-based alumni networking platform with 164+ commits. 4 dashboards (Alumni, Mentor, Coordinator, Admin), JWT auth, mock fallback, XSS/CSRF protection.
NLP-powered mood detection from free-form journal entries. VADER sentiment analysis with trend visualization, micro-anomaly detection for early burnout signals. Built at IBM India.
IoT-enabled smart bins with AI device identification, GPS tracking, weight/fill-level sensors, and real-time analytics dashboard. Mobile app for citizen engagement.
Multi-object detection & tracking with Python + OpenCV. Real-time video stream processing with bounding box visualization and class labeling.
Autonomous robot with PID control algorithm, Arduino Mega microcontroller, and 5-channel IR sensor array. Interrupt-driven motor control for real-time path detection.
SEO-optimized portfolio with dark mode, 3D animated particle backgrounds, JSON-LD structured data, GSAP animations. Responsive design with mobile-first approach.
National and global hackathon wins.
Led team to national victory — best solution from thousands of teams across India.
Predictive ML for water optimization. IoT + AI integration for precision agriculture.
Risk-based scoring & financial insights platform with AI portfolio recommendations.
Project, opportunity, or question? I'd love to hear from you.
Open to new opportunities, AI projects, or collaborations. I typically respond within 24 hours.