Soumya K.

Key Skills

Software

Roboflow

CVAT

Label Studio

Top Subject Matter

Autonomous Vehicles

Industrial Robotics

Reinforcement Learning

Top Data Types

Image

Top Task Types

Object Detection

Bounding Box

Polygon

Segmentation

Question Answering

Fine-tuning

Data Collection

AI Trainer Overview

Research Intern – Cobot Based Vision Analytics System for Quality Inspection In Automobile Process Line. Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include PyTorch. Education includes Bachelor of Technology, Jalpaiguri Government Engineering College (2027) and Bachelor of Science, Medinipur College (Autonomous) (2023). AI-training focus includes data types such as Image and labeling workflows including Object Detection.

Labeling Experience

Research Intern – Intelligent Outdoor Unit Checker (iOUC): Computer Vision–Based Inspection of AC Outdoor Units on Production Line

ImageObject Detection

During my research internship, I contributed to building a computer vision inspection system for AC outdoor units in an industrial environment. My work included training and optimizing YOLO-based models for detecting and segmenting screws and stickers using multi-camera image data. I designed a labeling and inspection pipeline supporting both real-time and recorded-video modes for comprehensive defect detection. • Annotated and labeled image data for presence, screw, and sticker detection tasks. • Optimized YOLO models for multiple detection and segmentation use-cases. • Developed per-side defect verification and integrated automated pass/fail logic. • Built a multi-camera GUI interface for both real-time and offline inspection analysis.

2025 - Present

Research Intern – Intelligent Defect Detector (iDD): Computer Vision–Based Detection of Gear Markings for Transaxle Assembly Validation

ImageObject Detection

As a Research Intern, I developed and optimized a multi-stage computer vision system for detecting and classifying gear markings in transaxle assembly lines. The project involved image data annotation for ROI extraction, gear detection, and mark classification using deep learning models. I validated and fine-tuned the labeling models to enhance accuracy and efficiency for automated industrial inspection. • Annotated image data for ROI, gear component, and marking identification tasks. • Trained YOLO and custom deep learning models for real-time image labeling. • Implemented and validated automated gear defect detection against reference benchmarks. • Developed a GUI-based inspection platform for user-driven review and analysis.

2025 - Present

Research Intern – Cobot Based Vision Analytics System for Quality Inspection In Automobile Process Line

ImageObject Detection

As a Research Intern, I worked on training and benchmarking deep reinforcement learning algorithms for autonomous vehicle simulation in a vision analytics system. My responsibilities included using image data from simulated environments to achieve high accuracy in obstacle detection, goal-reaching, and collision avoidance deployments. I fine-tuned reward functions and optimized model performance for real-world, cobot-based industrial navigation. • Trained DRL models (DDPG, DQN, PPO, SAC, TD3) on large volumes of simulated driving and vision data. • Conducted extensive image-based evaluation and labeling for strategy assessment. • Achieved 98% accuracy in image-based obstacle detection and 99% in collision avoidance. • Selected optimal models for industrial cobot deployment in variable environments.

2025 - Present

Education

Work History