Projects

Selected work, described carefully and without confidential company details.

Receipt collection agent

Agentic workflow from transaction scanning to Slack outreach and receipt attachment.

Problem: Receipt collection is cross-system operational work where missing documents, user follow-up, and backend state need to stay aligned.
What I did: Focused on backend integration, agent orchestration, reliability, and cross-service coordination.
What I learned: Production agents need clear boundaries, durable state, human-readable audit trails, and graceful fallback paths.
Concepts: AI agents, Backend integration, Auditability, Fintech workflows
Status: Private — details intentionally abstract. TODO: add safe public details if appropriate.

Shared skills, agents, MCP configs, and standards for internal AI adoption.

Problem: Engineering teams need reusable patterns and guardrails before AI tools become reliable parts of daily delivery.
What I did: Organized shared conventions, reusable agent instructions, and practical enablement material for internal adoption.
What I learned: Enablement works best when it is close to the codebase and shaped around real team workflows.
Concepts: AI enablement, MCP, Engineering standards, Developer workflows
Status: Private — TODO: add public artifacts if any become shareable.

Infrastructure thinking for auditable agent actions across backend services.

Problem: Agent actions in regulated systems need identity, permissions, policy, traceability, and a controlled execution model.
What I did: Explored architecture patterns for routing agent actions through explicit actors and service boundaries.
What I learned: The hard part is less the model call and more the contract between agents, users, services, and governance.
Concepts: Agent architecture, Actor model, Governance, Platform engineering
Status: Exploratory / private — TODO: clarify public scope.

Payment verification, integrations, regulated workflows, and reliability work in fintech.

Problem: Financial workflows need dependable integrations, clear operational visibility, and careful handling of edge cases.
What I did: Worked on backend systems and integrations in a regulated fintech environment.
What I learned: Reliability, observability, and auditability are product features in regulated domains.
Concepts: Fintech, Backend systems, Payments, Reliability
Status: Private — TODO: add only non-confidential specifics.

Market, product, and technology exploration for tailored medical examination services.

Problem: Healthcare AI must fit clinical, operational, and regulatory constraints rather than sit beside them as a demo.
What I did: Researched product opportunities around tailored medical examination services and regulated AI workflows.
What I learned: Fintech lessons around trust, workflows, and auditability transfer well to healthcare and longevity AI.
Concepts: Healthcare AI, Longevity, Regulated domains, Product research
Status: Research — TODO: refine project direction and public notes.