🚀 “Building AI that actually does something useful — pushing models from localhost to enterprise production.”
Results-driven AI Engineer specializing in production-grade Agentic AI, Multimodal RAG architectures, and Computer Vision forensics. Experienced in deploying low-latency, scalable ML pipelines on Microsoft Azure, building end-to-end MLOps CI/CD systems, and optimizing CPU/edge inference for enterprise Document Intelligence.
- 🏢 AI Engineer at LOLC Technologies (Sep 2025 – Present)
- 🎓 BSc (Hons) Data Science (Candidate) — Coventry University (UK) affiliated with NIBM Sri Lanka
- 📄 Published Researcher on IEEE Xplore (Focus: Reliable, interpretable, and secure AI systems)
- 🛡️ Cybersecurity Certified — OPSWAT Academy (Critical Infrastructure Protection & Cyber Fundamentals)
- 📍 Based in Colombo / Moratuwa, Sri Lanka
- 🤖 AI-Powered Document Compliance Agent: Architected a multi-agent system automating audits against regulatory banking frameworks — slashing manual review time by 70% with 100% compliance coverage.
- 📑 Multilingual Document Intelligence Platform: Engineered an enterprise OCR extraction engine with plugin APIs, delivering 98% field extraction accuracy and cutting manual data entry costs by 60%.
- 🔍 AI-Generated Document Fraud Detector: Built a sub-second CV forensics engine detecting synthetic and digitally tampered records with 99.2% accuracy and 100% recall.
- 🎙️ Voice Sentiment & QA Analytics Platform: Built a multi-lingual LLM contact center dashboard that boosted QA throughput by 40% and identified churn signals in real time.
- 🏦 Bank Statement & Portfolio Analyzer: Automated transaction ledger parsing across 300+ page multi-bank statements for instant querying and fraud risk profiling.
| Agentic AI & LLMs |
|
| Computer Vision & DL |
-8A2BE2?style=flat-square){kind=spacer}
|
| MLOps, Cloud & Big Data |
-E25A1C?style=flat-square&logo=apachespark&logoColor=white){kind=icon}
|
| Backend & Web APIs |
|
| Databases & Vector DBs |
|
| Security & AI Governance |
|
| Project | Description | Architecture / Stack |
|---|---|---|
| 🛡️ ClarityGate | Upstream circuit-breaker gatekeeper for e-KYC documents (NICs, Cheques, Passports) intercepting blur and specular glare before OCR. Sub-400ms CPU inference. | Python, ONNX Runtime, OpenCV, FastAPI, MCP Server |
| ✍️ signature-validator-matcher | Lightweight, production-grade biometric signature detection and verification engine for banking KYC pipelines and legal document authentication. | PyTorch, Computer Vision, Deep Learning, FastAPI |
| 🍃 TeaVision AI | 2-stage deep learning pipeline for tea leaf disease segmentation (91.2% sensitivity) and classification (96.1% acc) paired with a LangGraph RAG treatment advisor. | EfficientNet-B3 + SCSE, LangGraph, ChromaDB, PyTorch |
| 🏥 Hospital Triage Agentic System | Autonomous Supervisor-Worker multi-agent platform for real-time emergency clinical intake, triage level routing, and medical guideline synthesis. | LangGraph, Mistral (Ollama), Streamlit, Python |
| 🎭 Ko Mask MLOps Pipeline | End-to-end production MLOps pipeline for face mask detection with automated CI/CD retraining, drift monitoring, and Dockerized inference APIs. | YOLOv8s, MLflow, DVC, FastAPI, Docker |
| ⚡ TweetPulse Pro | High-throughput distributed streaming platform processing live Twitter sentiment analytics across containerized microservices. | Apache Kafka, Apache Spark (PySpark), MongoDB, Django |
| 🌴 SerenDipity Insights | Interactive analytical platform modeling Sri Lanka's tourism trends, economic indicators, and seasonal demand patterns. | Python, Dash, Plotly, Pandas |
- Published Researcher — IEEE Xplore
- Introduction to Critical Infrastructure Protection (ICIP, 2026) & Cybersecurity Fundamentals (2024) — OPSWAT Academy
- Elements of AI (2026) — University of Helsinki
- Professional Data Science with Python (2023) — University of Colombo School of Computing (UCSC)
- Executive Committee Member — Data Scientist’s Club, NIBM (2024 – Present)
- Member — SIAM (Society for Industrial and Applied Mathematics) Student Chapter
Designed with ❤️ by Manula Fernando • Let's build something extraordinary.