Work Experience
Associate Full Stack AI Engineer
Edge Computer Vision & Industrial Object Tracking (SaaS): Decreased inventory shrinkage by 40% and manual oversight by 80% by spearheading an edge-to-cloud pipeline containerized using Docker (C++17, Ultralytics YOLO model, AWS) achieving 30+ FPS real-time detection.
Chairman Saheb App Chatbot (SaaS): Accelerated user onboarding time by 30% by designing an interactive Flutter and NodeJS conversational interface with an integrated AI Agent that dynamically navigates applications.
Taparia Tools Manufacturing and Packaging Optimization: Improved operational efficiency by 25% by streamlining workflows with NVIDIA cuOpt prescriptive analytics and Azure ML for cloud-scale mathematical modeling.
Identity Verification & Anti-Spoofing System (SaaS): Engineered a deep learning-based presentation attack detection pipeline using OpenCV, FaceNet architecture, and TensorFlow, deployed via Docker and Android, achieving 92-95% accuracy in identity verification.
Vehicle Detection & Automatic Number Plate Recognition (ANPR) System: Trained and deployed custom convolutional neural networks leveraging PyTorch, TensorFlow, OpenCV, Ultralytics YOLO model, and OCR for real-time license plate recognition, exceeding 90% accuracy.
Physiotherapy Pose Detection: Created a real-time exercise monitoring application using Google MediaPipe, OpenCV, and React, providing angle calculations and form correction with over 95% landmark accuracy.
Collaborated with cross-functional teams to architect and deploy 5+ enterprise AI/ML models into scalable Software as a Service (SaaS) production environments using Agile methodologies, supporting over 10,000 daily API requests.
Led the end-to-end development of 4 major computer vision and Natural Language Processing (NLP) applications, managing scalable AWS cloud infrastructure to drive a 30% increase in overall system throughput.
Data Science & Research Intern
Formulated an Adaptive Kalman Filter (MLE), improving navigation accuracy by 37% and lowering noise by 30%.
Executed simulations for Moving Window Average (87% accuracy) and pioneered a hybrid Global Navigation Satellite System (GNSS) (Least Squares + Random Forest) achieving 4.48m positional accuracy.
Personal Projects
Diamond Defect Detection & Analysis: Decoded proprietary formats using Python, OpenCV, and reverse engineering to build a custom computer vision pipeline, saving $50K+ annually.
Trading Automation System: Improved trade execution latency by 15% by architecting a Python algorithmic platform that integrates ML predictive models and APIs for systematic risk management.
AI Agent - Conversational Assistant: Constructed a custom automation and customer support tool utilizing Python, FastAPI, and Natural Language Processing (NLP) driven intent recognition, decreasing manual interactions by 40%.
DOT Seismic Detector: Processed Apollo and Mars data via Python-based signal processing and data analysis techniques (noise filtering and quadratic fitting), achieving 96% event detection accuracy.
Education
Bachelor of Engineering (B.E.) in Artificial Intelligence & Data Science
Skills
Publications
Enhancing GPS Positioning Accuracy Using Machine Learning Regression
International Journal of Technology Engineering Arts Mathematics Science
Vol. 4, No. 1, June 2024, pp. 01~06, ISSN: 2583-1224