Principal engineer with 12 years building production AI and enterprise platforms. I ship MCP tooling, model routing, RAG, agent orchestration, and full-stack products across regulated environments, with measurable improvements to reliability and delivery. I am especially interested in aerospace and space infrastructure.
I build the engineering that turns models into dependable products: typed MCP tools, retrieval, model routing, evaluation, guardrails, and multi-step agent workflows. The result is production software that can use real context, take bounded actions, recover from failure, and remain understandable to the people operating it.
I generate and ship Model Context Protocol servers that expose existing systems as safe, typed tools, like the IMS MCP I built to bring AI tooling to parity with a core insurance platform's API surface.
Multi-provider setups across Claude, OpenAI, Gemini, and local Ollama, with fallback chains and per-task selection tuned for cost, latency, and reliability instead of one model for everything.
Designing multi-step agent workflows: planning, tool calls, and orchestration, with the guardrails and observability that make them dependable in production. The thinking behind FleetRun.
Retrieval, vector memory, and document intelligence embedded where teams already work, from in-browser vector memory to NLP chatbots and Microsoft 365 document automation.
Eleven custom Model Context Protocol servers give my agents typed, governed access to the tools a real software business runs on. The models run on hardware I own and route per task, so cost and latency stay under control.
AI lesson planning for K-12. Plans, worksheets with answer keys, and differentiated ELL/IEP versions from one prompt, aligned to all 50 states plus Common Core & NGSS.
Bible study & daily devotionals. A cross-platform web + mobile experience (Next.js 16 + Expo) for faith communities to study, reflect, and stay consistent.
Governed AI operations for solo operators and small teams: bounded missions, human approval before consequential actions, connected tools, readable traces, and visible run costs.
A quieter way to fall asleep. Original sleep stories, guided meditations, and a seamless soundscape mixer narrated in soft, real voices, now live on iOS and Android.
Outside my product work, I explore local AI, voice interfaces, and operating-system engineering. Selected code is available in my public repositories.
A browser-local recreation of "Her": speech-to-speech LLM with in-browser vector memory. ~2GB of models, zero server, fully private.
A voice assistant with 1940s British charm. Wake-word, Silero VAD, Whisper STT, and a plugin skills system: a voice layer over Claude CLI.
A from-scratch Unix-like OS: 25k+ lines, custom kernel, TCP/IP stack, VFS, and a macOS-style GUI. Runs ARM64 & x86_64, boots Doom.
A retro AI terminal inspired by Alien: CRT aesthetics, 7 themes, multi-provider (Ollama / OpenAI / Gemini / Claude) with a sci-fi knowledge DB.
A multi-source live stream aggregator that brings all your favorite streams into one unified place. Built with Next.js.
FleetRun, Kovu, NativeBlox, brain, MCP servers & more. Explore them all →
2014 → now · enterprise SaaS across healthcare, finance, legal & insurtech
Leading agentic AI and the API platform behind a modern insurance stack. Shipped IMS MCP servers with 30+ typed tools across entities, producers, financials, and reports, plus multi-provider routing with fallback chains tuned for cost, latency, and reliability.
My independent web and mobile studio, building calm consumer apps and trustworthy AI systems across faith, sleep, learning, and governed operations: FaithWise, Dreamtide, Taleglow Reading, FleetRun, and Nexani. leonenkogroup.com
Leading engineering and technology strategy as co-founder.
Azure and Power Automate platform work, plus Microsoft 365 / SharePoint consulting.
Modernized legacy healthcare systems into microservices and built SharePoint Framework solutions across Azure and REST APIs.
Where it started: full-stack product engineering on PHP & SQL for a camp management platform.
Principal AI and platform engineer with 12 years designing, building, and scaling production software across insurance, healthcare, legal technology, and independent product companies.
My strongest work sits at the intersection of AI systems and product engineering: TypeScript and Python services, React interfaces, APIs, MCP tooling, retrieval, model routing, evaluation, and the operational controls that make agents reliable.
I am looking for Principal or Staff scope where I can own architecture and still build, with aerospace and space infrastructure as the priority. I bring regulated-industry discipline, founder-level ownership, and a long record of turning ambiguous problems into shipped systems.
A public recommendation from a colleague at Northwell Health
“I'm happy to recommend Andrew Leonenko. I worked with him and was consistently impressed by his skills and dedication.