At Bayseian, I developed customer-facing AI applications integrating Python and TypeScript services, LLM workflows, and cloud deployments. One project was a Google ADK multi-agent marketing application with project-scoped memory and identity-aware image tagging.
I also developed an AI presentation-generation system that lets users edit specific content and layout elements with natural language, without regenerating the entire presentation. For job recommendations, I implemented semantic matching workflows using embeddings, cosine similarity, DBSCAN, skill extraction, and seniority-aware ranking.
At hSenid Biz, I developed enterprise HR REST APIs using C# and ASP.NET Core, and resolved React/.NET reporting defects and browser compatibility issues. I also developed a scheduled Python utility for production database loads.
In my AI/ML projects, I built Cognify AI, an agentic learning and RAG platform, and fine-tuned XLM-RoBERTa for Sinhala/Singlish PII detection. I also fine-tuned Gemma 3 4B using domain-adaptive pre-training, supervised fine-tuning, and LoRA.
