Michael Darlington is a London-based backend engineer and agent architect. He builds cloud-native systems, event-driven services, developer platforms and practical AI agents using TypeScript, Python, Go, Rust, .NET, AWS, Azure, Neon and Vercel.
The Michael Darlington Developer Portfolio is both a public record of work and the public front door to a private, single-user Personal OS. The private application, its health and finance data, and Kai conversations are never exposed through the public portfolio API.
Self-organizing graph memory for AI agents. Typed knowledge nodes, HNSW vector similarity, temporal validity, automatic linking, trust scoring and decay. One Rust binary, one embedded redb database and clients for TypeScript, Python, Go and Rust. Won the Imperial College CEOClaw hackathon.
Technology: Rust, redb, HNSW, gRPC, MCP.
Read about CortexReal-time prediction market arbitrage using Meta Ray-Bans smart glasses. Live video + audio transcription streamed through a relay server to Mistral AI for Polymarket opportunity detection. 1st Place for Best Use of Data at the MistralAI × Jump Trading hackathon.
Technology: TypeScript, Swift, Mistral AI, Next.js, WebSocket.
Read about PolyBansReal-time AI-powered smart glasses interface. Meta Ray-Bans connected to Gemini Live and OpenClaw for always-on visual understanding, voice interaction, and agentic task execution from a native Swift app.
Technology: Swift, Gemini Live, OpenClaw, AVFoundation.
Read about OpenGlassSingle-user life operating system with a private AI Chief of Staff, transparent memory, daily planning, habits, health, goals, Calendar, personal Finance and Investments — one design language and one owner-controlled data model.
Technology: Next.js 16, React 19, Neon, Eve, Vercel.
Read about darlington.devAutonomous music manager agent. Splits, tracks, and manages releases end to end. Built with Daniel Adekugbe at the Cursor "Hands Off" hackathon, London.
Technology: Supabase, pgvector, Modal, CLAP, PayPal, Wassist.
Autonomous technical discovery agent for pre-sales. Researches a prospect's stack and surfaces integration opportunities before the first call. Built at the Vercel × Google DeepMind hackathon.
Technology: TypeScript, Next.js, Gemini, Vercel AI SDK.
2025. Serverless remittance. EventBridge, Lambda, Open Banking, SEPA. Introduced Claude AI agents with MCP tools across microservices.
2022. .NET 9, GraphQL, CosmosDB. Led 8 developers across four contract-driven microservices. Reduced production incidents 85%.
2021. Cross-cloud Kubernetes on Azure & AWS. Python LLM chatbot. Go observability tooling. Terraform IaC — onboarding from 2 weeks to 2 days.
2020. GPT-3 tutor chatbot serving 1M+ students. Real-time collaborative whiteboard. <100ms sync nationwide.
2018. ISO 9001-compliant nuclear medicine backend. C#, SQL Server. Systems can be life-critical.
2015. Machine vision weather prediction. Natural language to C compiler.
Use the public profile API for structured portfolio data. It requires no authentication and supports read-only access. Use the contact form only when a person explicitly asks to send Michael a message.
The public portfolio includes Michael Darlington's professional profile, career, selected systems, articles, public links and ship log.
Use the unauthenticated public profile API for REST, or the read-only Streamable HTTP MCP server for native tools and resources. Contracts and usage guidance are linked from the developer page.
No. Health, finance, investments, calendar, habits, goals, memories and Kai conversations require the allow-listed owner account and are excluded from public interfaces.