About
Working philosophy
I came to AI through operations, not hype.
I have a background in finance, accounting, operations, and process improvement — and I build agentic systems that also reach beyond finance into commerce, verification, browser agents, and live assistance.
I naturally see organizations as systems: information comes in, decisions are made, work is routed, exceptions occur, controls are applied, and outputs move somewhere else.
AI introduces an entirely new kind of worker into that system. My focus is figuring out how to use it responsibly and practically.
I work across the gap between the business problem and technical implementation. I can map a workflow with the people performing it, design an agentic architecture around it, work directly with repositories and APIs, build and test the implementation, diagnose where it breaks, and iterate until the system becomes useful.
I am particularly comfortable when the process is still messy. Sometimes the hardest automation problem isn't automation at all. It is determining what the process should actually be.
The useful part is the overlap.
I operate in the translation layer between business ambiguity and technical systems — fluent enough in operations to know what actually matters, and hands-on enough in AI engineering to build it.
Business
01- Operations
- Process improvement
- Controls
- Commerce & workflows
- Finance (domain depth)
- Reporting
System design
02- Workflow mapping
- Automation architecture
- Data flow
- Failure modes
- Human controls
AI engineering
03- LLMs
- Agents
- Orchestration
- Tool use
- APIs
- RAG / context
- Verification
- Open protocols
- Observability
- Deployment
Operating Principles
How I decide what to build, and how to build it.
- Start with the work, not the model.
- The newest model does not matter if the workflow is wrong.
- Give agents jobs, not personalities.
- Clear responsibilities and contracts are more useful than elaborate personas.
- Evidence beats confidence.
- A system should be able to show why its answer deserves to be trusted.
- Automate decisions carefully.
- The more consequential the action, the stronger the verification and approval boundary should be.
- Build for whoever comes next.
- Documentation, observable behavior, tests, ownership, and failure recovery belong in the product.
- Ship the loop.
- A small workflow operating end to end is more valuable than a massive architecture that is 80% connected.
Resume
Download
AI Automation portfolio résumé — focused on agentic systems, automation leadership, and production AI operations work.
Download PDFTechnology
What I actually work with.
Models · IDEs · Agent Runtimes
- Claude (Anthropic)
- Claude Code
- OpenAI GPT
- OpenAI Codex
- Google Gemini
- Cursor
- ChatGPT
- Grok
- DeepSeek
- Mistral / Mixtral
- Llama / open-weight models
- Ollama
- LM Studio
- OpenRouter
- Azure OpenAI
- Amazon Bedrock
- Vertex AI
- Hugging Face
- LangChain / LangGraph
- LlamaIndex
- Vercel AI SDK
- Anthropic API
- OpenAI API / Assistants
- Function calling · tool use
- MCP (Model Context Protocol)
- A2A agent protocols
- Multi-agent orchestration
- RAG / embeddings / vector search
- Prompt libraries · eval harnesses
- Computer use / browser agents
- Voice · realtime audio pipelines
Engineering
- TypeScript · JavaScript
- Python
- Node.js
- Next.js · React
- FastAPI · NestJS · Express
- REST · GraphQL · Webhooks
- gRPC · WebSockets
- PostgreSQL · MySQL · SQLite
- MongoDB · Redis
- Prisma · Drizzle · SQLAlchemy
- Playwright · Puppeteer
- Tauri · Electron
- pytest · Vitest · Jest
- Docker · Compose
- Git · GitHub · CI/CD
- OpenAPI · JSON Schema
- Stripe · payment rails
- Ed25519 · hashing · audit trails
Infrastructure · Cloud · Ops
- Vercel
- Azure
- AWS
- Google Cloud
- Render
- Railway
- Fly.io
- Cloudflare
- Docker · containers
- Background jobs · queues · workers
- Cron · schedulers
- Sentry · logging · metrics
- Datadog / OpenTelemetry-style observability
- Auth (OAuth · JWT · SSO patterns)
- Secrets vaults · env management
- DNS · custom domains · TLS
- Supabase · Neon · PlanetScale
- Blob / object storage
- CDN · edge functions
Business Systems · Integrations
- QuickBooks Online · QuickBooks Desktop
- Xero
- NetSuite (integration patterns)
- Stripe · Stripe Connect
- PayPal · ACH / bank rails (ops patterns)
- Gmail · Google Workspace
- Outlook · Microsoft 365
- Slack · Microsoft Teams · Telegram
- HubSpot · Salesforce · Pipedrive
- Notion · Confluence · Google Docs
- Airtable · Sheets · Excel
- Zapier · Make · n8n
- Plaid-style financial data patterns
- DocuSign / e-sign workflows
- Twilio · SMS · WhatsApp Business
- Calendly · booking systems
- Metabase · Looker-style reporting
- SharePoint · Dropbox · Drive
- Jira · Linear · Asana
- Webhook / ETL / inbox → ledger pipelines
- CRM attribution · commission workflows
- AP / AR · invoice · expense systems