I build scalable backends & the agentic AI that runs on them.
I'm Theekshana Sandaru — a senior software engineer with 8+ years building scalable backend systems and AI platforms across big tech, fintech, and high-growth startups. I'm currently leading agentic AI architecture at MyTelescope: LLM-driven systems, multi-agent orchestration, and custom MCP servers. Based in Stockholm, open to senior and lead roles across the EU.
Lately, I build AI agents that actually ship.
For the past three years my work has centred on applied LLM and agent systems — from a retrieval assistant over a sovereign fund's financial documents at GIC, to leading agentic AI architecture today at MyTelescope. Not demos: production systems with real reliability, evaluation, and cost constraints.
Multi-agent orchestration
Coordinator/worker systems where specialised agents plan, delegate, and hand off — designed to stay controllable and debuggable, not magic black boxes.
MCP servers
Custom Model Context Protocol servers that extend Claude and other assistants with proprietary data, tools, and workflows.
RAG & retrieval
Retrieval-augmented systems over real corpora — embeddings, vector search, and grounding that keeps answers honest about what they know.
Agentic systems, end-to-end
LLM integration, tool use, and evaluation wired into reliable backend services — from the schema to the streamed response.
What I do
// expertiseAgentic AI & LLMs
LLM-driven systems, multi-agent orchestration, RAG pipelines, and custom Model Context Protocol (MCP) servers that extend assistants like Claude with real data and tools.
Backend at scale
Services and APIs in Python, TypeScript, and Go — Django, FastAPI, Litestar, Node — with the reliability that petabyte migrations and sovereign-fund workloads demand.
Data & platform
ETL and pipelines (Airflow, SFTP), Postgres and MySQL, and cloud-native delivery on GCP and AWS with Docker, Kubernetes, and CI/CD.
Stack
// tools of the tradeThe path
// since 2017ML-backed backends
Python APIs serving machine-learning model outputs at scale (Deep Data Insight).
Product engineering
Full-stack delivery across food-tech, climate-tech, and telco for AU/UK clients (iTelaSoft, Ayra).
Fintech + first AI systems
Analytics platforms and a RAG assistant over internal financial documents at GIC.
Big-tech scale
Petabyte-scale migration at Meta — USD 2.4M saved.
Leading agentic AI
Multi-agent orchestration and custom MCP servers powering MyTelescope's intelligence layer.
Experience
// selected rolesMyTelescope
· Lead EngineerLeading agentic AI architecture for an AI-native brand & market intelligence platform.2025 — PresentStockholm02
Meta
· Full-Stack EngineerPetabyte-scale document migration across Meta's storage estate.2024 — 2025Singapore03
GIC
· Full-Stack EngineerInvestment analytics and a RAG assistant over internal financial documents at a sovereign wealth fund.2022 — 2024Singapore04
iTelaSoft
· Software EngineerProduct engineering across food-tech, climate-tech, and telco for clients in Australia and the UK.2020 — 2022Sri Lanka05
Ayra Analytics
· Software EngineerHigh-performance software for the Octopus BI analytics platform.2020Sri Lanka06
Deep Data Insight
· Software EngineerPython backend APIs serving complex machine-learning model outputs at scale.2018 — 2020Sri Lanka
Earlier: Trainee Software Engineer at Vitte and Open Arc Systems (Odoo ERP), Sri Lanka, 2017.
BSc Information Technology — University of Moratuwa (2015 — 2019)
Final-year research: a card-fraud detection system using autoencoder neural networks (Python, Flask, Angular, Node.js).
Beyond the code
// off the clockSharing what I learn
Articles, videos, and talks on backend engineering, Rust, and building honest AI systems.
4 pieces →Let's build something that ships.
Open to senior and lead engineering roles across the EU — and interesting problems in backend and AI.