Full-Stack & AI Engineer
I turn business problems into shipped technical solutions.
Across both a product company and a services company, I own the full arc: understanding the problem, architecting the solution, and building it end to end, frontend through backend, data, and LLM orchestration.
Open to remote roles — US · Europe · Australia
Business problem in. Working software out.
- 01
Understand the problem
Sit with the business need, the users, and the real constraints. Write the requirements down — often a BRD — and get sign-off before any code.
- 02
Architect the solution
Choose the right shape: data model, services, where AI genuinely helps and where deterministic code must decide. Design for cost, safety, and change.
- 03
Ship & own it end to end
Build, test, deploy, and stay accountable for it in production. Requirements to running software, with no hand-off gap in the middle.
Selected work
Five builds where a business problem became running software. Each links to how it was built.
staging / demo data Circl8 Marketplace
A real, multi-user commercial product — shipped.
A B2B marketplace and reverse-logistics program to close the recycled-plastics loop — buyers, recyclers, orders, and analytics in one product.
- React
- Flask
- Typesense
- PostgreSQL
Business Audit Tool
Generative AI proposes. Deterministic code decides.
Grade a business's digital presence reliably — without letting a language model invent the grade.
226 passing tests
- Next.js 16
- Claude
- Prisma
- pg-boss
Self-Serve Chatbot Platform
URL in, live grounded chatbot out.
Let any business stand up a grounded support chatbot from just their website URL, with no engineering.
- RAG
- Pinecone
- Embeddable widget
- Lead-gen
Therapist Practice Platform
One practice, one platform.
Give therapists one place to run a practice instead of stitching together five tools.
- Next.js
- Supabase
- AI notes
Also built
Agent infrastructure, data pipelines, client sites, backend systems, and freelance apps.
-
Upwork Lead-to-Proposal Automation
An always-on, multi-account Gmail listener that turns inbound Upwork emails into Zoho CRM leads, sheet rows, and Cliq alerts — then auto-drafts AI proposals (Claude Haiku for quick, Opus for formal) and files them to WorkDrive.
- Python
- Claude
- Zoho / Gmail APIs
-
Zoho Agent Suite
A FastMCP agent server (~78 tools) with a 3-layer safety stack and human-in-the-loop preflight/plan/execute/rollback.
Case study- FastMCP
- Python
- OAuth
-
Content Automation Platform
Config-driven content generation (blogs, service & city pages) with a multi-provider model gateway and a closed-loop learning system — built, then retired when Alyne moved to a simpler Claude-Skills approach.
Case study- FastAPI
- Next.js
- Supabase
-
Web Agency Sites
Production Next.js client sites plus a design-token pipeline compiling one source into Tailwind, CSS vars, and WordPress themes.
- Next.js
- Design tokens
- Vercel
-
E-commerce Microservices
Spring Boot microservices with a JPA-backed OAuth2 authorization server, Strategy-pattern payments, Redis, and Flyway.
- Spring Boot
- Java
- Redis
-
Freelance — AI & Mobile
Two Flutter apps, a React web app for an AI podcast, and AI chatbots / voice agents built with Vapi and Voiceflow.
- Flutter
- Vapi
- Voiceflow
I've worked both sides: product and services.
At Etsaw, a product company, I built Circl8, a B2B recycled-plastics marketplace, starting on the React frontend and growing into backend services, search, and performance work. Shipping a real multi-user product taught me what it takes to keep one running.
At Alyne, a services company, I run client engagements end to end — turning business requirements into solutions the Zoho team implements, while I own the business analysis, solution architecture, client calls, team management, and ops; I go hands-on myself where the work needs integrations or custom code. I stood up the Zoho delivery practice from zero, shipped 7+ client projects through five team changes without a client-impacting delay, and was promoted to Tech Lead. Beyond client work, I built the automation that runs the business — an always-on Upwork lead-to-proposal pipeline, and an AI content platform Alyne later replaced with a simpler Claude-Skills approach. Now I'm integrating Claude — Skills, Projects, scheduled agents, MCP connected to our Zoho apps — into how the agency runs day to day, and leading CRM workflow development for the tech team.
The through-line is the same in both: I'm not only comfortable with the technical work, I'm comfortable sitting with a real business problem, understanding what's actually needed, and designing an intelligent solution for it, then building it. Lately that means production LLM systems, where I keep deterministic scoring separate from the model, ground outputs in real evidence, and engineer for cost and safety. I also get the most out of the AI coding tools themselves — building custom Skills, plugins, and MCP servers for tools like Claude Code so the tooling compounds my output.
Toolkit
Languages
- TypeScript
- JavaScript
- Python
- Java
- Dart
- SQL
Frontend
- React
- Next.js
- Astro
- Tailwind
- Flutter
- Design systems
Backend
- FastAPI
- Flask
- Express
- Spring Boot
- Workers / queues
AI / LLM
- Anthropic Claude
- Multi-agent pipelines
- MCP / agents
- RAG (Pinecone)
- Evals (Braintrust)
- Vapi / Voiceflow
Data / Infra
- PostgreSQL
- Prisma
- Supabase
- Firebase
- AWS
- Redis
- Docker
- GitHub Actions
Tools
- Claude Code
- Skills · plugins · MCP
- Cursor
- Git
- Playwright
- Postman
- JIRA