Backend depth.Productownership.
I architect backend systems at massive scale, and as founding engineer on a personal project, took a product from an empty repository to paying customers — alone. If you need someone who can own the whole build, that's the work I do best.
tokens/month through LLM infrastructure I architect
AI processing cost cut by moving inference in-house
SaaS platform: empty repo to paying customers, alone
years shipping production systems end-to-end
Three things, done properly.
Backend & Data Platforms
Python, Django, FastAPI — and the data layer underneath: Trino, Iceberg, ClickHouse, BigQuery, Spark. Systems that stay deterministic at a billion events a month.
AI Infrastructure
Production LLM pipelines, not API wrappers — self-hosted vLLM inference, prompt-scoped extraction, and the cost engineering that makes 5B tokens/month affordable.
Full-Stack Product
When it matters, I own the whole thing: Next.js frontends, Stripe billing, RBAC, GCP infrastructure, CI/CD — from architecture decision to the metric that proves it shipped.
The systems behind the numbers.
AI Hyper Cube — LLM Processing Infrastructure
The core LLM infrastructure that scopes and runs prompts over AI-generated search content at massive scale — extracting intent, brand entities, sentiment, and citations for enterprise SEO analytics.
Read the case studyTrino–Iceberg–ClickHouse Data Pipeline
Replaced a legacy BigQuery UDF workflow with a self-managed Trino–Iceberg–ClickHouse pipeline, cutting monthly processing cost by a substantial five-figure sum while scaling to over a billion events per month.
Read the case studyHigh-Throughput Keyword Collection Pipeline
A rebuilt keyword-collection system spanning multiple vendor integrations, achieving 4x faster processing at 100M+ keywords per month.
Read the case studyHealthcare Claims Intelligence Platform
A production SaaS platform that automates the collection, validation, and analysis of insurance claims data for healthcare organizations. I was employee/contractor number one — no codebase, no architecture, no team. I built all of it, then hired the engineers who followed.
Read the case studyA short history.
Building something that
needs to actually work?
I'm open to senior backend, platform, and full-stack roles — and to founding-engineer engagements where the product doesn't exist yet.