Automation engineering for businesses that are tired of copy-paste — web scraping, data pipelines, API integrations and AI agents, built and maintained end-to-end.
30k
SKU pipeline shipped
10+
Tools delivered in 2026
~90%
Manual effort removed
Over a decade working with US companies — a multi-year agency engagement, a Washington-state SaaS, and Upwork clients nationwide.
Overlap with US Eastern and Pacific working hours — standups, calls and same-day responses.
Source code, documentation and a handover walkthrough on every project. No lock-in, ever.
Contract and pay in USD through Upwork (100% Job Success), or direct with an NDA when you need one.
AI-assisted scrapers that survive layout changes — Playwright rendering, structured output, respectful rate limiting.
ETL from APIs, sheets and databases into clean, joined, scheduled datasets — the 30,000-SKU pipeline is the reference case.
Shopify, GA4, ad platforms, CRM and internal APIs wired together so systems talk without manual exports.
LangGraph multi-agent workflows, MCP servers and prompt pipelines with validation and human-in-the-loop gates.
Scheduled GA4 and ad-platform pulls into dashboards your team actually reads — no more Monday screenshot ritual.
Streamlit and web apps that turn a spreadsheet ritual into a one-click workflow with an audit trail.
STEP 01
Define the workflow, the data involved and the success metric — in one call.
STEP 02
A working proof on a sample of your real data, fast.
STEP 03
Error handling, logging, scheduling and monitoring added.
STEP 04
Docs, code ownership and optional maintenance retainer.
Recent work includes a Python pipeline that localized 30,000 product SKUs at ~90% less manual effort, an AI vision scraper and lead-generation MCP server, LangGraph multi-agent systems, and analytics dashboards deployed to Streamlit Cloud. All of it is documented in the AI Lab.
Scope-dependent — a focused automation script is a fixed-price project; ongoing pipelines suit a monthly retainer. Most clients start with one high-ROI workflow and expand.
Scraping public data is generally lawful, but terms of service, robots.txt, rate limits and data types all matter. I build within those boundaries and flag grey areas upfront.
Yes — I overlap Eastern and Pacific hours and have 12+ years of remote work with US companies.
Either a maintenance retainer with monitoring, or a documented handover so your team owns it. You always get the source code.
Python (pandas, FastAPI, Playwright, LangGraph, Streamlit) plus JavaScript/Node for browser tooling. AI work is model-agnostic — OpenAI, Anthropic, Gemini or Bedrock per job.
Describe the manual process — I'll tell you if it's worth automating, honestly.