Computer Vision ML Engineer with PyTorch

BHUPENDRASINGH

702-***-**** ***********************@*****.*** Pilani, Rajasthan, India 333031

SKILLS

Languages : Python, Java, C++, SQL

Core Concepts : Deep Learning Fundamentals, Computer Vision, OOP, DBMS Frameworks & Libraries : TensorFlow/Keras, PyTorch, OpenCV, NumPy, Pandas, Matplotlib EXPERIENCE

Devaditya Technocrats LLP, Indore, MP Apr 2026 –Present Software Engineer

Contributing to the development of a MLFF tolling system by integrating ANPR with Radar-based vehicle tracking for automated vehicle identification and toll transaction processing.

Trained YOLO-based vehicle detection and vehicle color classification models while designing and implementing a custom multi-head RT-DETR architecture in PyTorch for simultaneous vehicle detection, vehicle type classification (7 classes), and vehicle color classification(10 classes) in a single forward pass.

Deployed and validated the AI pipeline at Jogipura Toll Plaza, the first MLFF trial deployment site, by integrating synchronized ANPR and Radar outputs into structured JSON data, enabling real-time multi-sensor vehicle information processing and system validation under live traffic conditions.

Managed the NVIDIA GPU environment on Linux systems, including driver installation, CUDA/cuDNN configuration, and GPU validation to enable efficient model development.

Integrated and deployed trained computer vision models across ITS frameworks, including ANPR, AVC, VIDES, ATCC, and VASD, to enable real-time traffic monitoring and analytics. AI/ML Intern CSIR-CEERI, Pilani, Rajasthan June 2025 – Dec 2025

Developed an end-to-end ANPR system using YOLOv12 (PyTorch) for license plate detection and a custom CCT model for OCR- based character recognition on a 50k-image in-house dataset, with image preprocessing handled using OpenCV.

Trained YOLOv12 in PyTorch and exported the model through best.pt ONNX TensorRT (.engine) for high-performance, real-time inference.

Designed, trained, and evaluated the CCT-based OCR model using TensorFlow, achieving 98.6% character-level recognition accuracy on the validation dataset, and converted the trained model to ONNX to enable faster and more efficient inference.

Manage live ANPR operations across multiple toll plazas with real-time inference and continuous data streaming, while developing the CSIR-CEERI ANPR infrastructure with FTP services, REST APIs, and a Streamlit-based monitoring dashboard. PROJECTS

Vehicle Speed Estimation System

Built a real-time vehicle speed detection system in Python using YOLOv11, OpenCV, and Supervision to detect and track vehicles in video streams.

Designed an algorithm to calculate vehicle speed from frame displacement and FPS, ensuring accurate measurements.

Automated CSV logging of vehicle IDs, speeds, and timestamps for structured analysis. Generated annotated output videos for visualization of traffic movement.

EDUCATION

B.TechComputerScienceEngineering,CGPA:7.28 Nov 2021 – Jul 2025 BK Birla Institute Of EngineeringAndTechnology,Pilani Class-12th Percentage : 65 Apr 2019 – May 2020

Birla Shishu Vihar, Pilani

CERTIFICATION

Deep Learning Specialization – Coursera June 2025

Completed a 5-course specialization covering foundational and advanced concepts in deep learning.

Gained hands-on experience with neural networks, optimization algorithms, and TensorFlow-based implementations.

Built and trained models for computer vision, NLP, and sequence prediction, applying CNNs, RNNs, and LSTMs.

Applied best practices in deep learning workflows, including regularization, hyperparameter tuning, and model evaluation. Data Analytics – Udemy Oct2025

Gained hands-on experience with Excel, SQL (advanced queries & scenarios), Python, and Power BI for data analysis and visualization.

Worked on end-to-end data analytics workflows including data cleaning, quality checks, and structured reporting using real-world datasets.

Developed analytical dashboards and insights using Power BI, with exposure to data warehousing concepts and cloud tools such as Azure and Databricks.

Contact this candidate