randika@ai:~$ echo "Let's build something!" 🚀
|
|
name: Randika Prabashwara
current_role: Research Engineer @ Expert Intelligence™ (Nov 2025 – Present)
prev_role: Research Assistant @ University of Moratuwa (Jul 2025 – Feb 2026)
education: B.Sc. Engineering (Honours) — Computer Science & Engineering
University of Moratuwa, Sri Lanka | Specialization: Data Science & Engineering
focus: Computer Vision · Generative AI · MLOps · Production ML Systems
publications: 2 (IEEE ICITR 2025 + ADScAI 2025)
location: Sri Lanka / Remote
status: Open to research collaborations & exciting ML opportunities- 🔬 Research — Multi-scale super-resolution with GANs, PINNs & diffusion-based probabilistic models
- 🏭 Industry — Production AI systems at Expert Intelligence™ (international, remote)
- 📝 Publishing — GSM-SR paper in IEEE ICITR 2025; DSFRA paper at ADScAI 2025
- 🧠 Exploring — Multi-modal learning, RAG systems, financial AI
|
📄 IEEE ICITR 2025 GSM-SR: Generalized Structure-Informed Multi-Scale GAN for Efficient Image Super-Resolution Gayani Wickramarathna, Oshadi Perera, Randika Prabashwara, Uthayasanker Thayasivam 2025 10th International Conference on Information Technology Research · Paper ID: ICITR2025-145
|
📄 ADScAI 2025 A Data-Driven Spatiotemporal Framework for Retail Analytics Randika Prabashwara, H.K. Gayani V.L. Wickramarathna, Uthayasanker Thayasivam Applied Data Science & Artificial Intelligence Symposium 2025 · University of Moratuwa Library
Key results: 86.5% prediction accuracy · 20.3% delivery cost reduction |
Final Year Research Project · Jun 2024 – Oct 2025 · Team of 3 · Principal Researcher
Pioneering super-resolution framework for microscopy and nanoscience imaging. Novel architecture integrating encoder-based feature conditioning, multi-scale discriminator design, and a custom Semantic Structural Loss (SSL) function.
| Metric | Result | Baseline Improvement |
|---|---|---|
| PSNR (Set5/Set14/BSD100) | State-of-the-art | +3.5 dB |
| SSIM | 0.978 (SEM dataset) | +20% |
| LPIPS | 0.178 (SEM dataset) | −20% |
| Outperforms | EDSR, ESRGAN, SwinIR, HMANet | ✅ All benchmarks |
Python PyTorch TensorFlow Keras CUDA OpenCV Custom GAN
Brown & Company PLC · Dec 2024 – Oct 2025 · Lead Data Scientist
Enterprise-grade sales analytics platform translating a vague requirement ("better field visibility") into a concrete ML system: GPS-based route reconstruction, geospatial territory clustering (DBSCAN/OPTICS), and predictive opportunity scoring.
| Impact | Value |
|---|---|
| Sales Efficiency | +35% |
| Operational Cost Reduction | −40% |
| API Latency | < 500ms |
| Prediction Accuracy | 88% in production |
Python FastAPI Apache Kafka PostgreSQL Redis Docker Azure React
Independent Research · Aug 2025 – Present · Lead Developer
RAG-powered medical chatbot with multi-model LLM consensus routing (Llama, OpenAI, Gemini, Claude), Pinecone vector database for semantic search, and persistent conversation memory.
- 80% query classification accuracy across 4 integrated LLM providers
- Semantic search over medical knowledge base with source attribution per response
- Intelligent routing: model selected per query type (diagnostic / informational / drug interaction)
LangChain FastAPI PyTorch Pinecone OpenAI API Docker CUDA
📂 View all 9 projects
Apr 2025 – May 2025 · Independent
Multi-model OCR pipeline combining TrOCR, EasyOCR, and Enhanced Tesseract with 4-factor orientation detection, overcoming the 60–70% accuracy ceiling of existing solutions.
Results: 96% vehicle class accuracy · 92% date extraction · 98% orientation detection · 1.8s avg processing
TrOCR EasyOCR PyTorch OpenCV Transformers CUDA
Brown & Company PLC · Dec 2023 – Jun 2024 · Team of 3
3-layer OCR pipeline using Google Cloud Vision API, Azure Computer Vision API, and custom vision models for Sri Lankan bank cheque processing. Eliminated manual data entry bottlenecks with mobile banking integration.
Google Cloud Vision Azure CV Python OpenCV FastAPI
Brown & Company PLC · Dec 2023 – Jun 2024 · Independent
Fully automated data pipeline replacing manual Excel-based weekly reporting. Scheduled extraction from PostgreSQL, automated EDA with anomaly flagging, dynamic visualization, and templated report assembly delivered to stakeholders automatically.
Python pandas Matplotlib Plotly PostgreSQL Jinja2 Azure
University of Moratuwa · Jan 2025 – Jun 2025 · Teaching Assistant
Automated evaluation engine for RPAL programming assignments. Compiles, executes, and scores 100+ student submissions against comprehensive test suites with structured per-student feedback generation.
Impact: 70% reduction in grading time · 100% consistent scoring
Python Compiler Design Automated Testing AST Analysis
Mar 2025 – Jun 2025 · Big Data Project
Fully Spark-native distributed ML pipeline for music genre classification. Ensemble of Random Forest, Gradient Boosting Trees, and Logistic Regression with parallel feature extraction across distributed clusters.
Apache Spark PySpark MLlib MLflow Python Distributed Computing
Core Languages
ML & Deep Learning
MLOps & Infrastructure
Cloud Platforms
Data & Databases
|
|
|
|
🏅 Academic
|
📜 Certifications
|
|
🏭 Research Engineer
|
🔬 Computer Vision Research
|
🤖 Generative AI
|
📝 Publications
|
"Transforming complex data into intelligent solutions that drive real-world impact."
