AI Architect · Production agent systems
Systems that
think carefully
and ship cleanly.
I design and build AI systems and autonomous agents for teams that need reliability, not demos. Tool loops, multi-agent orchestration, and the integration patterns that turn model capability into software you can trust.
7+
Years shipping
15+
Systems in production
500K+
Users impacted
About
An architect who still writes the hard parts.
Blueprints that survive contact with production. Hands-on when it matters. Clear about tradeoffs either way.
How I work with clients
I translate business problems into technical blueprints: agent architectures, data flows, API contracts, and phased build plans with explicit tradeoffs.
Requirements discovery comes first, then deliberate architecture, then disciplined execution with AI-assisted engineering. Working systems, not just diagrams.
What I bring into a team
The same rigor, hands-on: design docs, model and infrastructure decisions, and shipping. I specialize in tool loops, multi-agent orchestration, model selection, and the integration patterns that make LLMs reliable software.
- Architected and shipped agentic systems: Oak and Annie ACS
- Enterprise AI deployed to production for clients across healthcare, finance, and recruiting
- 3 iOS applications published to the App Store
- Co-founded a game studio whose first title passed 500,000 downloads
Years experience
Systems shipped
Users impacted
System
Anatomy of an agent system
How I decide, then the seven layers I design before a demo. Each layer is grounded in Oak or Annie ACS.
How I decide
- An agent only where a workflow can't. Deterministic code is cheaper, faster, and testable; agents have to earn their place.
- Multi-agent only when isolation pays. Splitting context costs coordination, and the split has to buy more than it costs.
- The smallest model that passes the evals. Model choice is an eval result, not a brand preference.
- The harness before the feature. Evals and traces come first, so every change after is measurable.
Cross-section of a production agent system
Cross-section of a production agent system
Layer 01
Orchestration
The control flow above the model: plain workflow, single agent, or multiple agents with scoped context. Handoffs, retries, and stop conditions are designed, not emergent.
In practice
Annie ACSPatient path is route, retrieve, respond, not a free-form free-for-all. Insights is a second, scoped agent for staff.
Selected work
Case studies with the architecture left in.
Flagship systems up front. Everything else as a clean index. Outcomes and decisions, not marketing copy.
Oak
A full-stack AI agent that answers natural-language Pokémon questions by reasoning over raw game data, not just looking facts up, with cited sources and explicit uncertainty in every answer.
14
Reasoning tools
3
Drop-in LLM providers
TypeScript · Next.js · Postgres · Drizzle · Server-Sent Events · Zod
Read case study
Annie ACS
An AI medical consultant for a cosmetic surgery practice that educates patients on procedures, streams cited media from a curated knowledge base, books consultations via Calendly, and gives practice staff a full admin panel, including a second agent that answers natural-language questions over thousands of patient conversations.
0
Hallucinated media URLs
3
Embed modes
Python · FastAPI · React · MUI · Vite · Postgres
Read case study
More work
Nibble AIAn AI-powered nutrition tracker that uses Claude AI for instant meal analysis from photos, available on the App Store.AI/ML2024LogoForgeA free AI-powered logo generator that creates App Store-ready icon bundles using Google's Gemini AI.AI/ML2026FamilyCartA collaborative grocery list app with real-time sync and zero-friction onboarding, built for families.Mobile App2024YosihealthA mobile app for digitizing and organizing health records, built for the Indian market where paper-based medical documents are still the norm.Healthcare Tech2023Suit BreakA multiplayer card game with real-time matchmaking, ELO rating system, and strategic gameplay mechanics.Web Game2024Laboratory Information Management SystemLed the development of a custom LIMS that replaced Salesforce, enabling a diagnostic lab to process 10,000+ samples daily.Healthcare Tech2022-2023DMEPOS Order Management SystemLed the development of an order management system for a DMEPOS provider, replacing manual paper processes with a streamlined digital workflow.Healthcare Tech2022-2023AI-Powered Expense Validation & AuditingDesigned an AI solution using OCI Generative AI to automate expense report validation and auditing for a large enterprise.Enterprise AI2025AI Resume ParserBuilt an AI-powered solution that bulk-filters 1000+ job applicants and returns top candidates with detailed analysis based on configurable criteria.Enterprise AI2024ParaverseLed product development for a blockchain mobile app bridging physical and digital worlds through NFT-based redemption experiences.Web3/Blockchain2021-2022Agent REDAn autonomous multi-agent AI system that plays Pokemon Red using Claude AI models and a Game Boy emulator.AI/ML2024Capabilities
Where depth lives.
Not a laundry list of tools. The competencies that show up when systems have to work under real constraints.
Contact
Let's build something that holds up in production.
Open to architecture engagements, full-time roles, and conversations about agentic systems that need to be reliable.
Email me