I build.
Others
just talk about it.
From zero to production. AI systems, dev tools, and data platforms, shipped, not pitched.
5+ years turning complex problems into products people actually use.
Self-taught.
Obsessed with shipping.
I don't do concept phases. Problem in, product out, short cycles. Dev tools, AI infrastructure, data systems, built to sell, not to demo.
Ship first. Polish in production.
Technology I
actually use.
5+ years.
Production-grade.
AI Product Engineer
Independent
Multiple production systems designed and shipped end-to-end. LLM-powered agents, document processors, and prompt pipelines, all running with real users, real load, real constraints.
Numbers don't lie.
Real metrics from real production systems.
Query batching + Redis cache
Model routing: agile models for speed, robust models for accuracy
AI pipeline automation across workflows
Active projects, heavily adopted by open-source communities
What I shipped
that actually runs.
One product in development. Five tools open-sourced.

FlipperZ
Research-to-document agent. Turns a question into a professional deliverable where every claim is verifiable down to the statement level.

LicensePulse
OSS license watchdog. Monitors your dependencies and alerts you when a license changes, before it becomes your legal problem.

LeadStrike
B2B prospecting engine. Finds companies with high reputation but weak digital presence and delivers the lead ready to close.

Promptinel
Drift monitor for LLM pipelines. Catches silent behavior changes in your prompts before your users do.
PublishCanary
Scans your npm tarball before publish. Blocks secrets, source maps, and internal files before they go public.
LLMs are not hype.
They are infra.
RAG pipelines, multi-model routing, memory layers. Not demos. Production systems handling real latency, real token limits, real users.
Less wrapper. More infra.
Conversational Agents
Proprietary · Open Weights
Document Processing
Automated extraction + structuring
RAG Pipelines
Embeddings + vector search
Prompt Engineering
Structured pipelines
On-Device Inference
TFLite · ONNX -70% Cost
LLM Orchestration
Multi-step LangChain