Available for high-impact engineering & research collaborations
Senior Software Engineer / MSc in Data Science & AI (reading)
I design and build scalable, cloud-native platforms: distributed backends, enterprise tooling, and AI-powered services built on LLM orchestration, RAG and vector search. Currently leading engineering at H2O.ai, and researching how to make AI-generated code secure.
I'm a Senior Software Engineer with 6+ years designing and building scalable, cloud-native and enterprise platforms. My core is backend engineering and distributed systems (Go, gRPC, PostgreSQL, Kubernetes and AWS), with full-stack reach across TypeScript and React when a product needs it end to end.
At H2O.ai I own mission-critical platforms, drive architecture decisions, and build AI-powered capabilities: multi-provider LLM integration, RAG pipelines, vector search, and streaming APIs deployed on Kubernetes. I care about systems that are secure, observable and genuinely production-grade.
Alongside engineering, I'm reading for an MSc in Data Science & Artificial Intelligence at the University of Moratuwa, where my research tackles the security of LLM-generated code through automated red teaming.
M.Sc. in Data Science & Artificial Intelligence
Reading. Research on security of LLM-generated web application code.
B.Sc. (Hons) Electronic & Telecommunication Engineering
GPA 3.55/4.2, Second Class Upper. FYP: safety enhancement of pedestrian crossings.
H2O.ai
H2O.ai
Dialog – University of Moratuwa Laboratory
Applova Inc.
Production platforms and products I've designed, owned and shipped.
An AI-powered code generation service that translates natural language into executable code and configuration across multiple products and languages. Built end to end with streaming LLM responses, RAG over product documentation, vector retrieval, authentication, and Kubernetes deployment.
A production-grade, cloud-native platform for managing and administering environments, covering user management, security, billing tracking, observability, and identity management across enterprise deployments.
An enterprise-grade internal analytics platform that tracks telemetry analytics across all environments, giving teams a single lens on usage and system health.
An internal access-control platform built from scratch. Team members request, approve and track time-limited access to EKS clusters and AWS accounts with role-based permissions, real-time notifications and a full audit trail, solving a real infrastructure-security problem.
A one-month forward deployed engagement for a SaaS product built for the plantation sector. I helped frame the product with stakeholders, shaping operational needs into practical features and workflows, and coordinated the needful with the field teams across solution design, implementation and rollout.
A Red Teaming Approach to Characterising the Functional-Security Correctness Gap
The area. Large Language Models such as GPT-4o, Claude and DeepSeek-Coder are now widely used to generate web application code. My research sits where AI meets application security: it studies how safe that generated code really is once it becomes a complete, deployable web application rather than an isolated snippet.
The gap I'm addressing. Code an LLM produces can pass every functional test and still be insecure. Studies report that around 68.8% of generated snippets break security rules, while the share of real-world output that is both functional and secure stays below 6%. I call this divergence between "it works" and "it is safe" the functional-security correctness gap, and my work characterises it through adaptive red teaming so that AI-assisted development can be guided by evidence instead of assumption.
Public work and research code. More on github.com/SDPcodes.
A proof-of-concept from my MSc research. GPT-4o generates full-stack apps under different prompting strategies, then each is run through a security pipeline: functional test, static analysis (Semgrep, CodeQL, LLM-as-judge), sandbox deployment and red teaming (OWASP ZAP plus scripted attacks).
An AI agent that autonomously searches the web and synthesises grounded answers to a user's query.
Using Apache Spark to classify music by genre, with a train-then-serve ML pipeline.
Competed as Team PSR and placed 8th in the final round of H2O.ai's Wildfire Challenge, building an AI application end to end with H2O Wave. Started from scratch and learned a huge amount through the journey. Teammates: Ranush Wickramarathne and Pasan Dharmasiri.
Generative AI Foundations
Machine Learning for Natural Language Processing
Machine Learning Foundations
Data Engineering
H2O AI Agents: From Theory to Practice
Fundamentals of Accelerated Data Science
h2oGPT Intro Course
LLM Studio Overview
Driverless AI Starter Course
Encoder-Decoder Architecture
What colleagues and mentors have said on LinkedIn. Scroll for more.
Shamil was a highly motivated and talented intern who contributed effectively to our team for six months. He displayed strong dedication, ensuring high-quality and timely delivery of assigned tasks. Eager to learn, he proactively collaborated with engineers across the RF, Data Com, and Core teams, gaining end-to-end exposure to mobile network components. His willingness to take on new challenges and self-driven attitude made him a valuable team member. I highly recommend him for future opportunities.
I had the opportunity to work with Shamil for one year at Dialog, University of Moratuwa Research lab. We worked on several projects involving 5G technology and wireless communication. Shamil is always focused and determined to finish the allocated task before a given time frame, achieving the best results. He is a valued team member, and his dedication and ability to face issues with a positive attitude is remarkable for any team. Shamil will be a valuable asset for any company.
I've had the pleasure of working with Shamil on multiple projects. He's a curious and quick learner. In addition, he's easy to trust, since he takes on responsibilities well.
I've worked with Shamil on multiple telecommunication projects and he is well versed on the subject. A fast learner and committed to complete the scope undertaken.
Shamil is one of the easiest people to work with. Very hardworking and a dedicated worker.
Open to engineering roles, research collaboration, and consulting on backend, cloud-native and AI platform work. The fastest way to reach me is email.
Say hello