PHOTO CAPTION AGENT
Multi-agent AI pipeline that generates captions from photos — vision model + story direction + refinement loop, running via Ollama Cloud.
HELLO, I'M
NADAV SOMEKH
Full stack engineer for founders shipping their first product.
Web apps, APIs, data and AI features, built end to end. Oakland, Bay Area, California.
WHAT I DO
WEB APPS & APIS
Custom web apps, APIs, and backend systems, built to keep working past launch.
DATA LAYER
Database design and optimization, so the product stays fast as your data grows.
ECOMMERCE
Storefronts, checkout, and payment integration.
MVP BUILDS
From idea to a product your first users can sign in to, as an MVP build or a founding engineer engagement.
AI FEATURES
LLM integration and AI-powered features that do real work inside your product.
TECHNICAL CONSULTING
Technical advice on your product, architecture, or stack when you need an experienced engineer's input.
FRACTIONAL ENGINEERING
Part-time engineering work when you need experienced help without a full-time hire.
Skills
SYSTEM ARCHITECTURE & API DESIGN
Designing modular, scalable systems across REST, GraphQL, microservices, and event-driven architectures. Experienced breaking apart distributed monoliths into clean service boundaries.
BACKEND DEVELOPMENT
Node.js, NestJS, Express, Python, FastAPI, TypeScript. Deep expertise in authentication and authorization (JWT, OAuth 2.0, RBAC, API security).
FRONTEND DEVELOPMENT
React, Tailwind, MUI, responsive design. Strong UX/UI instincts with a focus on performance and usability.
DATABASES
PostgreSQL, MongoDB, Redis. Schema design, migrations, query optimization, and caching strategies.
CLOUD, CONTAINERS & DEVOPS
AWS, GCP, Docker, Kubernetes, GitHub Actions, CI/CD pipelines.
AI ENGINEERING
LLM integration, multi-agent workflow design, prompt engineering, vector search and embeddings, local LLM deployment with Ollama.
Demos
Examples, POCs, and samples from github.com/Nadavys
EVENT SOURCING WITH KAFKA + FASTAPI
Pure event-driven microservices built on event sourcing — services communicate exclusively through Kafka topics, no direct calls.
MULTI-AGENT RESEARCH PIPELINE
Implementation of a multi-agent AI pipeline — the pattern that produces better outputs than asking an LLM a single question.
MUSIC STORE MICROSERVICES
Full-stack microservices showcase — NestJS + Python workers + RabbitMQ RPC + LGTM observability stack, built on the Chinook music dataset.
MOVIE QUERY AI
Ask your database questions in plain English — Claude turns them into safe, read-only SQL and visualizes the results.
FULL-STACK GRAPHQL MOVIE RATINGS
React 19 + Apollo + Prisma monorepo for movie ratings, with end-to-end type safety shared across frontend and backend packages.
Shipping your first product? Tell me about it.
START A PROJECTFAQ
Who do I usually work with?
Founders shipping their first product, and anyone who needs technical consulting or fractional engineering work.
What's the difference between an MVP build and a founding engineer engagement?
An MVP build takes your idea to a product your first users can sign in to. A founding engineer engagement depends on the startup, so we agree on hours and length together.
What kind of technical consulting do I offer?
On the frontend: architecture, UI/UX, and SEO. On the backend: architecture, API design, scalability, and observability.
How much does it cost?
I quote after we've talked through what you're building, because the cost depends on scope. Send a short note and we can start there.
How long does an MVP take?
It varies with what the first version has to do, how many outside systems it connects to, and whether there's existing code to work with. Tell me about any deadline in your first note.
Why hire an engineer when AI can build everything?
AI is great for getting a prototype working fast, but it soon stops scaling and bugs start to pile up. To stabilize a product and help it grow, you need someone with real technical knowledge.
Do I take on existing prototypes?
Yes. I review what you have first, then recommend whether to keep and harden it or rebuild it. That includes AI-generated prototypes.
Do I work with non-technical founders?
Yes, happily. I explain decisions in plain English and keep you involved as the product takes shape.
Can I add AI to an existing product?
Yes. I build LLM integrations and AI features that do real work inside your product. For an example, Movie Query AI turns plain-English questions into safe, read-only SQL.
Do I work remotely?
Yes. I'm based in Oakland, CA and work remotely with founders worldwide.
How do we get started?
Send a short note about what you're building, where it stands, and any deadline. I'll reply by email with questions and, if it's a fit, a time to talk it through.