KUGANRAJ
SELVARAJ
AI-focused Software Engineer with hands-on experience in Computer Vision, AI model deployment, and full-stack software development within manufacturing environments. Strong background in anomaly detection, classification, image preprocessing, REST APIs, and CI/CD workflows.
// 01Core Skills & Capabilities
Programming & Software Development
AI / Machine Learning / Vision
DevOps & Version Control
Vision Systems & Automation
Tools & Platforms
// 02Professional Experience
Dev Ops Engineer – AI & Software Development
Oct 2024 - PresentJabil Penang
- Developed real-time Python systems to capture MQTT data, convert Base64 image payloads into structured JSON and image formats for inspection pipelines.
- Designed and implemented RESTful APIs (FastAPI) for secure data exchange, token validation, and controlled API access across systems.
- Built a Windows Forms (C#) application to centrally manage Raspberry Pi devices and robotic arm.
- Implemented Retrieval-Augmented Generation (RAG) using Ollama LLM (Gemma3), LangChain, and ChromaDB to enable document-based Q&A systems.
- Deployed AI services via FastAPI, hosted on IIS with defined backend/frontend ports for enterprise-wide access.
- Built full-stack dashboards using Next.js (TypeScript), amCharts, pdfjs, ExcelJS, and MySQL for analytics and reporting.
- Managed source code using Azure Repos, enforced branch-based development, conducted code reviews, and merged via pull requests following CI/CD best practices.
- Collaborated with Process and Industrial Engineers to validate software releases and align AI solutions with manufacturing workflows.
- Mentored 4 interns in software engineering, AI workflows, and version control standards.
- Exploring vision system integration with Omron robotic arms using Sysmac Studio, EtherCAT communication, and PLC coordination.
Intern – Process Engineer (Software & Automation)
Mar 2023 - Sep 2023Flex Penang
- Delivered 3 production-enhancing software tools within 6 months, improving operational efficiency.
- Designed and deployed an internal web dashboard for management-level production monitoring.
- Recognized with Best Intern Award for outstanding solo project contributions and problem-solving impact.
// 03Education
Bachelor of Computer Science (Hons)
2020 - 2024Universiti Sains Malaysia (USM)
- Major: Artificial Intelligence | Minor: Mathematics
- Gold Award – Final Year Project
Pre-University / Foundation
2019 - 2020Penang Matriculation College (KMPP)
- CGPA: 4.00 (Pure Science)
// 04Key Projects
AI Vision Anomaly Detection System
- Developed inspection pipelines using OpenCV and deep learning models for defect detection.
- Implemented noise reduction, exposure control, and threshold tuning to minimize false positives.
- Applied classification and region-based analysis to differentiate acceptable vs defective samples.
Compliance Document Q&A System (RAG + LLM)
- Built FastAPI-based AI backend serving document-aware responses using LangChain, ChromaDB, and Ollama.
- Implemented chunking, vector embeddings, and similarity search to ensure accurate, context-bound answers.
- Integrated frontend clients (Next.js, C#, Python) via robust API architecture.
USM Bus Buddy – Final Year Project
- Developed a real-time bus tracking and ETA prediction system using Google Maps API.
- Trained a neural network in Python, converted to TensorFlow Lite, and deployed to a mobile application.
- Awarded Gold Award for technical excellence and practical impact.
// 05Technical Glossary
Anomaly Detection
Technique used to identify patterns or images that deviate from normal conditions, commonly applied in defect detection for industrial vision systems.
Classification
Machine learning process that assigns labels (e.g., good/bad, defect type) to images or data based on learned features.
Keypoint Detection
Method for identifying specific points of interest in an image (edges, corners, features) used for alignment, measurement, or object localization.
Exposure Time
Camera parameter controlling light intake; optimized exposure improves image clarity and reduces false detections in vision systems.
Noise Reduction
Image preprocessing techniques (filtering, smoothing) used to remove unwanted artifacts that negatively affect model accuracy.
Azure Repos
Git-based version control service within Azure DevOps used for managing source code, pull requests, and collaboration.
CI/CD
Automated process that builds, tests, and deploys software to ensure faster, safer, and more reliable releases.
Branching Strategy
Version control practice where separate branches are used for features, testing, and production to maintain code stability.
EtherCAT
High-speed industrial communication protocol enabling real-time data exchange between controllers, vision systems, and robotic arms.
Sysmac Studio
Omron’s integrated development environment used to program PLCs, motion control, robotics, and industrial automation systems.