Nikodem Dabski

I started as a data analyst at McKinsey in Wrocław, took an internship at Swisscom in Bern, and found a second home in Switzerland. The thread through everything since: systems that hold under load, and teams that grow along the way.

Now. Technical Team Lead at Swisscom, with 9+ years in enterprise data, cloud, and AI. My team builds an agentic AI product that lets enterprise users query their data in natural language: text-to-SQL over a data mesh, with a semantic metadata layer as the agents' context, running on AWS and on-premise.

AI platform. I designed and built the service layer (Go and Python) behind Swisscom's inference endpoints: model routing, rate limiting, token budgets, and billing for hundreds of thousands of daily requests across LLMs, vision, and embedding models served through NVIDIA NIM. It powered a national AI initiative.

Data platforms. I ran JupyterHub as a service for hundreds of data scientists and analysts, and helped build Swisscom's in-house big data framework and the pipelines on it, moving data from the data warehouse into the Hadoop and Spark cluster behind the company's data lake.

Study. CAS in Advanced Machine Learning, University of Bern. Final project: open-source LLMs on text-to-SQL with agentic self-correction.

Originally from Poland, based in Luzern. I work in English and speak conversational German.

How I work

I focus on systems that work in production, not just in the demo, and on teams that come out of every project more capable than they went in.

Whether I'm writing, teaching, or mentoring, the goal is the same: making sure people understand the thinking behind the decisions, not just the decisions themselves.

On the side, I build Polycard, a flashcard app born from 10,000+ cards of certification prep. More about me at nikodemdabski.com.

What I work with

LLM / GenAI

LangChain, LangGraph, RAG pipelines, text-to-SQL, agentic workflows, prompt engineering, NVIDIA NIM, vLLM

Cloud & Infrastructure

AWS (12 certifications), Kubernetes (Kubestronaut), model serving, GPU infrastructure, Infrastructure as Code

Engineering

Golang (5 years), Python (9 years), microservices, event-driven architecture (Kafka), Prometheus/Grafana monitoring

Leadership

Distributed teams (multi-site), hiring, architecture decisions, Scrum Master experience, learning community organiser

Selected achievements

Swiss AI Weeks, national scale

The inference platform I built served Swiss AI Weeks at 99.997% reliability under peak load: 430,000+ requests and 250M tokens for 1,400 API keys, with 2.98 ms P99 rate-limiting overhead.

Recognition

Swisscom Learning Award: 1st place, Team category (2024), for the department learning community; 3rd place, Individual category (2025). Selected for the internal Talent Program five years running (2022/23 to 2026/27).

Community & teaching

Co-founded a department learning community and drove it for four years: 40+ events with 50 to 200 participants each, and a team of five moderators.

Research

CAS Advanced Machine Learning (University of Bern). Research: open-source LLM text-to-SQL with agentic self-correction on the BIRD benchmark.

Education

CAS in Advanced Machine Learning · University of Bern (2025–2026)

Final project: open-source LLM text-to-SQL with agentic self-correction on the BIRD benchmark.

Postgraduate Studies in Project Management · SGH Warsaw School of Economics (2016–2017)

Poland's top-ranked economics university, listed in the Financial Times Top 100 European Business Schools.

MSc in Behaviour and Decision-making in Business · Wrocław University of Science and Technology (2015–2016)

A broad business programme: finance, accounting, founding a company, and working with people. The business half of the engineering-plus-business pairing.

BSc in Computer Science · Wrocław University of Science and Technology (2011–2015)

Engineering degree (inżynier), the Polish professional title for technical programmes.