I’m Pavindran Velalagan, a third-year Computer Engineering undergraduate at the University of Peradeniya. I like building systems that combine software, hardware, and machine intelligence.
Recently, I worked as an ML Engineering Apprentice at Cogntix, where I worked on document text extraction, CV classification models, and AWS cloud deployments. Before that, I worked as a casual instructor, helping undergraduate students with Python and Object-Oriented Programming.
Most of my projects sit somewhere between embedded systems, full-stack development, and applied AI. Some of the work I’ve done includes a closed-loop IoT medical device for monitoring and regulating IV drip flow rates, a desktop app for students to learn Image Processing in an interactive manner, and a published research paper(Q1 Journal) on Vision Transformers for tomato leaf disease recognition. I’m usually interested in problems where both engineering and problem-solving matter equally.
Outside academics and engineering work, I’m actively involved with the IEEE Computer Society. Before that, I was involved with the AerogenZ Aviation Club and the Perabeats Media Society, where I worked as a videographer, editor, and webmaster.
Beyond engineering, I also work as a freelance photographer, mainly covering portraits and events. What started as a hobby gradually turned into professional work, and I enjoy capturing people, moments, and stories through photography.
I’m always open to interesting projects, research collaborations, and good conversations. Feel free to reach out.
Tutorials, documentations, and deep dives on software, hardware, and LITERALLY WHATEVER!
ESP32 · MQTT · C++ · Hardware Design · IoT
Developing a closed-loop medical device to bridge the gap between inaccurate gravity-fed drips and costly volumetric infusion pumps. Implements an intelligent stepper motor mechanism with an IR drop counter to automatically regulate flow rate, and integrates IoT connectivity via MQTT to a centralized ward dashboard for real-time monitoring.
Flutter · Dart · SQLite · Cross-platform · Mobile App
Kanakku is a minimalist, offline-first Flutter app. Developed in spare time to manage shared expenses and split bills among friends, as well as track active subscriptions. Built with a clean offline-first architecture for easy local ledger keeping.
Python · PyQt5 · OpenCV · NumPy · Matplotlib
An offline desktop lab application for the Digital Image Processing course. Implements visualization for syllabus-aligned operations including Histograms, K-Means segmentation, Fourier Transform, and simple CNNs. Features a "Learn This" panel providing intuition, formulas, and real-world use cases alongside live code execution.
React · TypeScript · Node.js (Fastify) · PostgreSQL · Google Gemini
Full-stack API-first monorepo with role-based JWT authentication and Zod schema validation. Integrated an AI-driven "Budget Bot" powered by Google Gemini to analyze expenses and surface spending insights. Features automatic database bootstrapping, mock data seeding, and expense analytics dashboards.
Python (Django) · JavaScript · PHP · MySQL
Solo-developed a web app allowing CSV/Excel data import and generation of complex analytical graphs. Selected as a finalist in the Yarl Geek Challenge 2018.
Vision Transformers · Deep Learning · Python · AgriEngineering
Pavindran Velalagan et al. Published in MDPI – AgriEngineering, June 2025.
A robust ViT-based framework for automated recognition of tomato leaf diseases,
achieving high accuracy with a focus on practical agricultural deployment.
→ Read
paper
Awarded for Smart IV, our 3rd Year Project, at the SLASSCOM National Ingenuity Awards 2026.
Secured 1st Runner Up in the IoT competition with the Smart IV project.
Selected as a finalist in the IoT competition with the Smart IV project.
Selected as a finalist in the national IoT competition with the Smart IV project.
Co-authored ViT-RoT: Vision Transformer-Based Robust Framework for Tomato Leaf Disease Recognition, published in MDPI AgriEngineering (J.1). First listed author.
Selected as a finalist for the Intelligent Traffic Monitoring System (ITMS) with Team Zephyrus at the IESL Student Chapter competition hosted by the University of Moratuwa. Proposed an AI-driven solution with dynamic traffic signal control and emergency vehicle prioritization.
Selected as a finalist for a solo-built Data Analytics Web Application at the Yarl Geek Challenge hosted by Yarl IT Hub. This was back in my school days in Grade 11 : )
Worked on text extraction for complex multi-column CV layouts, AI-based CV filtering, document classification models, and cloud deployment using AWS (EC2, Route 53, Lightsail, Lambda).
Short thoughts, ideas, and whatever comes into my mind...