Founding Engineer. Building the trace-based evaluation engine and the MCP tool layer, CLI, and SDK that turn agent failures into verified fixes.
2026–Present
Founding Engineer · PhD in AI · Co-Founder
PhD in AI · Researcher · Founding Engineer
Chamath
Palihawadana
Dr. Chamath
Palihawadana
I’ve spent over a decade turning ideas into products people rely on, and I still enjoy every part of shipping them.
Building Glassray after leading applied AI at Vera. Over 10 years shipping products from first commit to production.
PhD in Artificial Intelligence from Robert Gordon University, focused on federated learning. I study how machines can learn from data without ever centralising it.
Read as
Now building
Founding Engineer
The reliability layer for AI agents: it catches the silent failures monitoring misses, then ships back a verified fix teams can trust.
Notable research
FedSim
Similarity-guided aggregation that makes federated learning faster and more accurate, published in Neurocomputing.
Current role
Founding Engineer, building the reliability layer for AI agents
2026–Present
Doctorate
PhD in Artificial Intelligence, Federated Learning
Robert Gordon University, 2024. Similarity-guided methods that make federated learning faster and more accurate while the data never leaves the device.
How I Build
Multi-Agent Systems
LangGraph orchestration with planning, routing, and tool selection, measured against real evals.
Retrieval & RAG
Grounding LLMs in private knowledge with retrieval pipelines built for accuracy.
Evaluation & Observability
Ground-truth datasets, benchmark runs, and tracing with LangSmith and Langfuse to keep agents reliable.
Developer Tooling
CLIs, SDKs, and MCP servers exposing 30+ tools, plus cloud coding agents that clear routine tickets.
Scalable Backends
Microservices and serverless on AWS and Azure, built to hold up under heavy traffic.
Integrations & Payments
Slack, Teams, and Stripe integrations, plus voice agents built on LiveKit.
Experience
Privasee Group · Founding Engineer across two ventures
2025–Present
Founding Engineer. Led the move from a single-flow RAG assistant to a LangGraph multi-agent architecture with 93%+ answer accuracy, and shipped Vera’s MCP server plus Slack and Microsoft Teams integrations. Moved within the group to found Glassray.
2025–2026
Xergy Group
Senior Software Engineer. Led full-stack development of a project progress reporting platform, carrying it from build through testing and optimisation to release.
2023–2024
Research Assistant. Built the iSee explainable-AI cockpit and CloodCBR, the world’s first microservices-based case-based reasoning framework.
2019–2023
Earlier roles: Associate Software Engineer at 99X Technology · Visiting Lecturer at IIT.
Ventures
Technical Co-Founder. Scaled the LMS to 50,000+ students and 50 TB of monthly traffic, integrated five payment gateways processing 200M+ in payments, and built multilingual voice agents with LiveKit and RAG.
2020–2023
Co-Founder. Low-cost Bluetooth attendance used by over 10,000 students. Named Most Innovative Product at the RGU Startup Accelerator, 2020.
2019–2023
Co-Founder and CTO. Led engineering across more than 20 projects serving over 30,000 users.
2014–2022
Research & Publications
PhD in Artificial Intelligence (Federated Learning), Robert Gordon University, 2019–2024. My research spans privacy-preserving learning, explainable AI, and case-based reasoning.
Doctoral research
My thesis improved federated learning on three fronts: similarity-guided model aggregation (FedSim), similarity-guided feature extraction and pruning, and defences against gradient-inversion attacks.
Federated Learning
Similarity-guided aggregation and privacy-preserving training.
Explainable AI
iSee, a toolkit for sharing explanation experience between users.
Case-Based Reasoning
CloodCBR, a case-based reasoning framework built as microservices.
Adversarial Robustness
Defending federated models against gradient-inversion attacks.
Selected publications
FedSim: Similarity-guided model aggregation for Federated Learning
C. Palihawadana, N. Wiratunga, A. Wijekoon, H. Kalutarage · 2022
JournalNeurocomputing 483
Mitigating Gradient Inversion Attacks in Federated Learning with Frequency Transformation
C. Palihawadana, N. Wiratunga, H. Kalutarage, A. Wijekoon · 2023
ConferenceESORICS 2023
Adapting Semantic Similarity Methods for Case-Based Reasoning in the Cloud
I. Nkisi-Orji, C. Palihawadana, N. Wiratunga, D. Corsar, A. Wijekoon · 2022
ConferenceICCBR 2022
A Comparative Study of Link Analysis Algorithms
C. Palihawadana, G. Poravi · 2018
ConferenceISMS 2018
Recognition: AWS Community Builder (Serverless), 2022 · Second Place, 3-Minute Thesis (3MT), RGU, 2022 · University of Westminster Prize for Outstanding Achievement · Member of the British Computer Society (MBCS).
Education
Doctor of Philosophy, Artificial Intelligence. Thesis: Improving Federated Learning Performance with Similarity Guided Feature Extraction and Pruning.
2019–2024
BEng (Hons) Software Engineering. Final-year research on finding experts in Q&A platforms.
2015–2018
Skills
AI & Agents
Multi-Agent Systems
RAG
MCP
Voice Agents
LangGraph
LLM Evaluation
Federated Learning
Engineering
TypeScript
Python
React / Next.js
React Native
FastAPI
AWS & Azure
PostgreSQL
Serverless
Writing
Notes on applied AI, multi-agent systems, and the craft of shipping products.
We put Claude on the night shift (opens in new tab)
How we built an overnight engineering shift: Claude routines pick up Linear tickets, open PRs with evidence attached, and revise until the automated review comes back clean.
MediumAug 2026 · 4 min read