I build intelligent AI agents and workflow automations โ from Python-powered multi-agent systems with CrewAI, to API-integrated pipelines in n8n, to prompt-engineered tools that ship fast and think on their own. I also evaluate LLM outputs and agent tool calls for quality, correctness, and reliability โ rating tool-calling responses, validating JSON schemas, and applying structured scoring frameworks on real annotation datasets.
I'm Milena โ an AI Engineer, AI Agent Builder, and Automation Architect who builds across the full automation stack. I write Python to orchestrate multi-agent systems with CrewAI, integrate APIs and design pipelines in n8n, and apply prompt engineering to build tools that ship fast and think on their own.
My background includes leading a team of AI trainers at Scale AI as one of the project team leads โ one of the world's leading AI companies โ giving me a rare combination of hands-on building skills and a deep understanding of how AI systems learn and perform in the real world.
Based in Nairobi, Kenya, I have a particular focus on the African AI ecosystem โ tracking how AI is reshaping industries across the continent and building tools that put that intelligence directly in the hands of African builders and decision-makers.
I believe prompt engineering is a superpower. You don't need to write every line of code to build something powerful โ you need to know what to build and how to make it think.
An agentic AI tool that ingests any course content โ raw text, learning objectives, or module summaries โ and instantly generates a fully structured, difficulty-calibrated assessment. No LMS required. Just a shareable link your team can access instantly.
View on GitHub โBuilt to explore how AI and automation tools can eliminate manual invoice processing. The pipeline watches Gmail for PDF attachments, uses Claude AI to extract structured invoice data, routes low-confidence extractions for human review, and delivers a weekly HTML summary email โ all without human intervention.
View on GitHub โAn end-to-end issue reporting automation built on Zapier. When a team member encounters a problem, a form submission triggers an automated chain โ logging to Google Sheets, sending a confirmation email to the reporter, and firing a Slack alert to the responsible stakeholder for immediate action.
View on GitHub โA 4-agent CrewAI pipeline that autonomously researches the latest AI developments across Africa โ startups, funding, policy, tools, and events โ filters results for relevance, generates a structured weekly briefing, and delivers it as a formatted HTML email via SendGrid.
View on GitHub โI map out the workflow, identify what needs to think and what needs to automate, and design the architecture โ whether that's a multi-agent pipeline, an API integration, or a no-code automation chain.
I write Python or configure agents in CrewAI for complex reasoning tasks, or build visual pipelines in n8n, Make, or Zapier for workflow automation โ picking the right tool for the job.
I wire up APIs, external services, and data sources to make the system functional end-to-end โ connecting search, email, databases, messaging, and AI models as required.
I run the full pipeline end-to-end, validate outputs, and deploy. Interfaces are built with Replit or Lovable when needed, and every system is refined until it runs reliably without intervention.
Open to freelance projects, collaborations, and conversations about AI agents and automation. If your team has a workflow that should be smarter โ let's talk.