|
I am an AI undergraduate at the University of Moratuwa, Sri Lanka. I care about turning messy real-world problems into systems that are accurate in the lab and reliable when deployed.
Currently exploring: scalable backends with FastAPI, interactive frontends with React, and CV stacks around YOLO and classical vision tooling. |
|
End-to-end style project focused on where things are and what happens when they move—suited for campuses, warehouses, or mixed indoor/outdoor sites.
+ Indoor + outdoor tracking with a unified mental model
+ YOLO-style object detection for reliable visual cues
+ WebSocket channels for real-time alerts to dashboards or clients
+ Geofencing rules (enter / exit / dwell) for safety and logistics
+ Designed as a real-time intelligent system, not a batch-only demoAI-assisted health risk awareness app: users enter metrics (e.g. age, BMI, blood pressure) and get early signals for conditions such as diabetes—built with the FARM mindset (FastAPI, React, MongoDB).
+ Full-stack workflow: API + persistent storage + UI
+ Clear input forms for structured health metrics
+ Model-driven predictions surfaced in the product layerStroke risk assessment from patient-style health parameters using deep learning, with emphasis on interpretable inputs and model iteration in notebooks.
+ Neural models trained on tabular / clinical-style features
+ Jupyter-centric experimentation and evaluation
+ Framed as an intelligent decision-support style pipelineRain vs. no-rain (or similar classification) from meteorological features—classic ML baseline through to stronger models, documented in notebooks.
+ Feature-driven ML on weather-related data
+ Notebook-first storytelling: EDA → model → metrics
+ Practical forecasting-style problem formulationAgent-based simulation of a smart microgrid in Python with Mesa: solar, wind, batteries, consumers, and mobile maintenance agents, used to study emergent behavior and resilience.
+ Multi-agent modeling (not a single monolithic simulator)
+ Heterogeneous actors: generation, storage, demand, maintenance
+ Visual layers that highlight coupling and “butterfly effect” style dynamicsNotebook- and pipeline-oriented work that pushes a data science project toward repeatable runs and deployable artifacts—bridging experiment and operations.
+ MLOps-flavored structure on top of DS workflows
+ Experiment tracking, packaging, or deployment-oriented steps (per repo)
+ Good fit for portfolio storytelling around production AI🔷 NOLIMIT
TypeScript project (fork / collaboration track) complementing the Python-heavy side of the profile—useful for typed frontends, tooling, or full-stack glue.
+ TypeScript codebase for typed, maintainable UI or services
+ Pairs well with FastAPI + React stacks in other repos
If you’re working on computer vision, real-time systems, or ML in production, feel free to reach out by email or open an issue on a relevant repo.