About Alex Beattie
I’m a Forward Deployed Engineer at McKinsey & Company, working on a QuantumBlack life sciences engagement where I build LLM and agent systems directly with the client team.
Before that I was an Agentic AI Software Engineer at Technia, building AI tooling and client-facing applications on top of PLM systems, and the sole AI and backend engineer at Ourself Health, where I architected Stella, a production AI assistant on AWS Bedrock. I focus on systems that are reliable, observable, cost-aware, and practical to operate after launch.
In AI systems, I care most about the pieces that make products dependable after the demo: routing, grounding, source metadata, Langfuse evaluation loops, Strands agent behavior, observability, and operational controls.
Recent work includes:
- KiNDD Resource Navigator: nonprofit app for families navigating developmental disability services in LA County, live on the App Store
- Software engineer on a UCLA and Dup15q Alliance team awarded one of Anthropic’s AI for Science rare disease research grants
- Unified Connectors MCP Server: IDE-native engineering context across GitHub, Slack, Jira, Confluence, Google Docs, and Miro
- Cost-aware model routing, token analytics, and budget-aware AI system design
Background
I have 14+ years of experience shipping software across mobile, backend, cloud, and AI systems. I started in Swift/iOS, moved through TypeScript and Node.js, and now work primarily in Python, Django, Postgres, AWS, and LLM infrastructure.
Before moving into healthcare AI, I spent 13 years as the sole technologist at Sherwood Development Company, where I owned software, automation, web systems, integrations, and technical operations end to end.
What I Write About
- AI/LLM Engineering: agent architectures, RAG design, eval harnesses, prompt systems, model routing, and production observability
- Full-Stack Architecture: Django, PostgreSQL, GraphQL, AWS infrastructure, streaming APIs, and mobile/backend contracts
- Senior Engineering Practice: debugging complex systems, making tradeoffs explicit, and turning production lessons into reusable patterns
Tools I Reach For
AI/Agents: RAG pipelines, MCP servers, multi-agent orchestration, LLM-as-judge evals, model routing, vector search, AWS Bedrock, Claude
AI Platforms: Langfuse, Strands Agents, Pinecone, Cohere, pgvector, Titan embeddings
AI Dev Tools: Cursor custom rules/skills, Claude Code, MCP servers, agentic development workflows
Backend: Python, Django, GraphQL, Go, Node.js, REST APIs, PostgreSQL
Cloud: AWS ECS, Lambda, Bedrock, Aurora, S3, Docker, GitHub Actions
Frontend & Mobile: Flutter, SwiftUI, Vue.js, React