I’m a team player and a fast learner. I focus on creating the best output to get the best outcome possible. As a full stack engineer with 6+ years of experience, I integrate fast into new teams through strong communication skills and an understanding of the importance of teamwork.

My work centers on money-moving systems: virtual USD accounts, USDC transfers, webhook pipelines, and ledger correctness under retries. I design scalable, event-driven architectures, keep high standards for code quality, and build modular microservices. I use typed functional programming in TypeScript so the code stays robust and testable, with clear rules for concurrency and data consistency across distributed systems.

I also build grounded AI agents and agentic payment flows. Models propose. Systems must not invent state, and they must not spend without policy. I use an AI-assisted development workflow (spec first, tests before trust, human-owned architecture). Give me a Hi!

• ✉️ You can contact me at josemfcheo@gmail.com
• 🤝 I'm open to collaborating on interesting projects that help people around the world

Backend & Frontend

• TypeScript, Node.js, Effect-TS, Express, NestJS, Ruby on Rails, Django
• React, Next.js, Vue/Nuxt
• GraphQL, Prisma, Zod

Databases & Cloud Architecture

• PostgreSQL, MongoDB, DynamoDB, SQL, NoSQL
• AWS (Lambda, SQS, SNS, S3, API Gateway), Vercel, Cloudflare, DigitalOcean
• Docker, CI/CD: GitHub Actions, general DevOps familiarity

Fintech & Stablecoins

• Virtual USD accounts, USDC rails, crypto↔fiat flows, KYC and banking integrations
• Ledgers, idempotency, on-chain confirmation vs mempool credit, webhook pipelines
• Agentic payments: HTTP 402 / x402, gasless USDC authorizations (EIP-3009 / Permit2), EVM settlement
• Policy-gated spend: allowlists, budgets, rate limits, human approval, decision isolation (agent proposes, signer executes)

AI & Agents

• Tool calling / function loops, RAG with citations, embeddings, vector search (pgvector)
• Fail-closed grounding, prompt-injection guards outside the model, eval harnesses (recall@k, faithfulness)
• Voice-agent pipelines: STT → LLM → TTS, barge-in / VAD concepts
• LangChain, LangGraph, Vercel AI SDK, OpenAI API, Anthropic Claude, Hugging Face

Developer Productivity

• Cursor, Claude Code, Codex
• Agentic Plan–Build–Test loops, AI-assisted refactoring, automated test generation
• Spec-first context, isolated verification, human-owned architecture and review gates

• Financial Integration & Cloud Architecture: Owned planning and execution of virtual USD accounts and transfer systems. Integrated KYC providers, crypto payments, and banking services on event-driven AWS (Lambda, SNS, SQS FIFO, DynamoDB). Enforced ledger credit only after on-chain confirmations, idempotent writes under retries, and typed financial boundaries so amounts cannot coerce across fiat and crypto.

• Agentic Payments & Stablecoins: Built agents that pay for resources with USDC over HTTP 402 paywalls. Separated the decision to pay from the key that signs: the model creates a structured intent; deterministic policy checks amount, allowlist, budget, and rate limits; a human or policy signer executes. Used gasless authorizations and on-chain settlement. Guards live in code, not in the prompt.

• Grounded AI Agents: Built fintech chat agents with function-calling tools and RAG over tool-derived facts plus citation IDs, instead of stuffing last-N messages into context. Fail closed when evidence is missing. Measured with offline evals, not prompt demos. Same rule as the ledger: do not invent state.

• Performance & System Optimization: Handled large datasets, optimized query performance, and built real-time third-party integrations via Node.js, GraphQL, and Prisma.

• Automated Testing & QA: Built automated testing frameworks with Node.js, Jest, Mocha, and Selenium. Reduced regression issues and manual testing time. Treat flaky or stochastic behavior as product risk, including in agent and eval work.

• SaaS & Tooling: Delivered full-stack features for SaaS platforms, real estate CRMs, and automated digital marketing tools. Translated non-technical requirements into clean, functional implementations using React, PostgreSQL, and Ruby on Rails.

• Production applied AI: grounding, tool use, evals, and agent safety (the model is untrusted; policy is not)
• Stablecoins and machine-native payments: USDC, micropayments, HTTP 402, agents that can pay without inventing spend
• Money systems where fast and correct are not the same: ledgers, retries, confirmation, and support-load as a design input
• High-growth financial markets and trading