Rahul Kumar - Deep Learning & Computer Vision Engineer
Innovative Deep Learning Engineer with expertise in computer vision and AI-driven solutions. Proficient in YOLO, PyTorch, and TensorFlow, with a strong track record of fast API-based application development.
Contact Me
Professional Experience
1
Deep Learning Engineer
Add Innovations Private Limited-Vision Systems (01/2024 – present)
Developing datasets for detection, segmentation, and tracking using industrial cameras. Expertise in Yolov5/8 and optimizing GPU performance with CUDA.
2
Computer Vision and Data Science
Algorizz Technologies Private Limited (09/2024 – 04/2025)
Developed AI-powered traffic management systems with 95-98% accuracy across AVC, ANPR, ATCC, and VIDS modules.
Core Competencies
Deep Learning
YOLOv5, YOLOv8, YOLONAS, Object Detection, Image Segmentation, Generative AI (LLMs, Transformers)
Computer Vision
OpenCV, Haar Cascade, OCR, X-ray Diagnostics, PaddleOCR, TrOCR
Deployment
Microsoft Azure, Google Cloud Platform, REST APIs, FastAPI, Uvicorn, Docker
Automated Real-Time Potato Inspection System
Developed a real-time potato inspection system using dual Basler cameras and YOLOv8 for instance segmentation and object detection. The system identifies potatoes, measures dimensions, and detects surface defects with high accuracy.
Technologies: Python, YOLOv8, FastAPI, PyTorch, OpenCV, Multithreading, Basler Cameras
Automatic Number Plate Recognition (ANPR)
Developed a real-time OCR-based system for number plate recognition with ≥98% accuracy, optimized for 30ms inference time. The system includes vehicle classification, brand detection, and violation monitoring.
Key Features: Real-time plate recognition, vehicle classification, brand detection, and speed/violation detection.
Bolt Classification and Detection System
Project Overview
Developed a real-time bolt inspection system for Tesla's production line using YOLOv8 and ONNX, achieving 95% accuracy in detecting and classifying bolts ("NORM" vs. "TVS").
Technologies
FastAPI, Uvicorn, OpenCV, YOLOv8, ONNX
Impact
Reduced quality control errors by 10% through non-conforming image logging and robust classification.
OCR-Based Image Processing System
Developed a real-time OCR system for industrial text extraction using the TrOCR model. Applied OpenCV and PIL for image preprocessing to optimize text readability in challenging industrial environments.
Implemented a FastAPI-based REST API, socket programming for device communication, and multithreading for efficient processing, with CSV logging for timestamps and serial numbers.
Real-time OCR for Production Lane
Developed a real-time OCR solution using a custom-trained PaddleOCR model, optimized with ONNX and OpenVINO, and integrated via FastAPI for laser marking. Enhanced efficiency and reduced downtime in production lanes.
Technologies: PaddleOCR, ONNX, OpenVINO, FastAPI
Technical Skills
AI Frameworks
  • Keras, TensorFlow, PyTorch
  • Hugging Face Transformers
  • YOLO, PaddleOCR, TrOCR
Data Analysis
  • Pandas, NumPy, Dask
  • Matplotlib, Seaborn
  • Data Preprocessing & Augmentation
Tools & Platforms
  • Git, GitHub, Jupyter Notebook
  • Google Cloud Platform, Azure
  • FastAPI, Uvicorn, Docker
Education & Training
1
Master of Computer Application
Shri Krishna University
2
Bachelor of Computer Applications
Chhatrapati Shahu Ji Maharaj University
3
Python, Data Science, AI Training
Ducat Institute (06/2023 – 12/2023)
Full-time training in Data Science, Machine Learning and Artificial Intelligence
Based in Noida, India. Available for deep learning and computer vision projects.
Made with