Pipeline Doctor AGENT
An agent that fixes failing test suites. It runs the tests, reads the failing code, writes a fix, then re-runs - looping until its own verifier sees the suite fully green.
$ whoami ▍
const ryan = new AI_Automation_Engineer();
I design and build AI agents and workflow automation that do real work for Kenyan businesses — answering customers, tracking stock, reconciling cash, chasing invoices — using n8n, Zapier, Make, Gumloop & OpenClaw.
$ whoami
Ryan Eromosele Osagiede — AI Automation Engineer
$ cat skills.txt
zapier · zap · gumloop · n8n · make · openclaw · supabase
$ ls ./projects
intervio/ ecom-support/ jobhunter/ shop-hunter/ pipeline-doctor/ document-extractor/ deep-researcher/ qa-agent/ +5 in the roadmap
$ ▍
The stack I build with — rendered the way I think about it:
Shipped, documented, and running — every project is a working system you can inspect, not a mockup.
An agent that fixes failing test suites. It runs the tests, reads the failing code, writes a fix, then re-runs - looping until its own verifier sees the suite fully green.
Extracts fields from messy invoices, validates the arithmetic, and fixes its own mistakes - recomputing totals and VAT until every document is clean.
Writes a report from its sources, then verifies every claim. Anything it cannot back up gets rewritten until the whole report is supported - no invented facts make it through.
Starts a live web app, runs test scenarios, and files bug reports with reproduction steps - refusing to finish until every failure is documented.
A memory-aware mock interviewer that asks one tailored question at a time and grades your answer — verdict, score, what was missing, and a model answer — then drills your weak spots across sessions. Runs locally in the browser and over WhatsApp via OpenClaw — a working live demo.
A WhatsApp-first AI support agent for Kenyan e-commerce vendors. An LLM classifies customer intent, the agent answers order-status, product and returns questions from the shop's own data, creates RMAs, and escalates to a human — designed to cut response time from hours to seconds.
Searches Kenyan job boards daily for 6 target roles, matches each posting to a role-tailored CV, tracks applications, and pings WhatsApp when something new shows up. Built to make my own internship hunt systematic.
Sends one honest WhatsApp message per day to approved shops and businesses offering workflow automation, AI agents, or websites — with an approval gate, a 5/day cap, and a permanent opt-out blocklist. Built to turn my automation skills into clients.
Live interactive traces - watch each agent plan, call tools, and verify its own work. Powered by DeepSeek, 4/4 tasks passing. Full lab and eval harness: github.com/ryanreo/agentic-workflow-lab.
Trading bots, health tools, and other projects outside the AI-agent world.
Software Development student at KCA University (2023–2026) and freelance web developer who fell into AI automation and never looked back. I build practical, low-cost, Kenya-ready systems — WhatsApp, SMS, USSD, M-Pesa — for the SMEs and informal-sector businesses that automation usually ignores.
My superpower: turning a business problem ("cashiers spend an hour counting the till") into a working, logged, human-approvable AI workflow.
I'm currently focused on landing an AI automation / agent development internship where I can take these systems from working demos to production.
const ryan = {
location: 'Nairobi, Kenya',
focus: 'AI agents for Kenyan SMEs',
stack: ['n8n', 'zapier', 'make', 'gumloop', 'openclaw'],
education: 'BSc Software Development · KCA',
status: 'building in public',
};
Have a workflow that's eating your team's time? Let's automate it.