About me
I'm Borbála Tasnádi, a mathematician turned freelance AI engineer based in Berlin. I help businesses take the step from AI curiosity to AI in daily work, and help tech teams make their GenAI hold up in production.
After studying mathematics at FU Berlin, I started out in data science in 2017 and then spent four years as an ML engineer in privacy tech. My work there centred on measuring data quality and protecting sensitive data. For the past three years I've focused fully on generative AI: I build knowledge bases, automate document processing and develop AI agents that work in real workflows.
Along the way I've learned that good AI projects start with the problem, not the technology. So every project begins with a short scoping phase: we identify your pain points and pin down the goal, the data and how we'll measure success. Then I build in small, tested steps: rules where rules are enough, AI where it actually helps, and a test set from your own data that shows how well it works. Where AI isn't reliable yet, I say so and design around it.
8+ years in AI, machine learning & data · 6 years building Python backends
What I offer
🧭 Scoping: we agree on goals, risks and whether AI is the right tool at all.
⚙️ Workflow automation & AI agents: AI steps built into your existing processes, so routine work is done and your team handles the exceptions.
📄 Document processing: invoices, contracts and forms read and processed automatically.
📚 Knowledge bases (RAG): one AI assistant across Slack, Confluence, Google Drive and your files, with every answer linked to where it came from.
✅ Quality & cost control: testing and benchmarking on your own data, so you get the quality you need at a cost that makes sense.
Every system I build is traceable and privacy-first: answers link to their sources, and personal data stays where it belongs.
My tech stack includes Python, LangChain, LangGraph, PydanticAI, FastAPI, Django, n8n, vector databases (Pinecone, Chroma, pgvector), Postgres, Docker, AWS, Azure, and GCP, with Langfuse and Phoenix (Arize) for LLM observability and evaluation.
-
Want to find out what AI can do for your processes?
Book a free 30-minute intro call. I'll ask a few direct questions and give you my honest take on what's possible and what it would take.