Ines Panker, senior staff software engineer and conference speaker. 16 years in software, the last 6 as the first engineer and de-facto architect of a SaaS observability platform. The engineering teams built on what I designed, and I led the response when complex incidents crossed system boundaries.
Lately my work has turned toward LLMs: I’ve worked on the LLM pipeline for our AI-usage analytics tool (a tool that gives insights into what your team does with AI). I’ve shipped my first RAG, written many Claude skills and rules and have built several agents for my own tooling.
I also love speaking at conferences, like Devoxx Belgium (3,000+ developers, sells out in minutes), J-Fall (the biggest Java conference in the Netherlands) and Python Italia. Some talks come straight from the systems above: what happens when a third-party API goes down, and how to tell if a deploy broke production. The others are about the lies we tell ourselves: how we estimate, how we argue, …
Python Go Django Flask Celery PostgreSQL Redis RabbitMQ Elasticsearch
Multi-tenant SaaS Circuit breakers Distributed locking
Pandas Anomaly detection Time series
LLM agents Context engineering RAG
Technical writing Documentation Teaching Conference talks
PHP MySQL JavaScript C# Java
The first engineer they hired, and the de-facto architect ever since. Sleuth is a multi-tenant SaaS observability platform for DORA metrics, environment health and real-time anomalies, fed by 30+ third-party services. I joined when it was two coding founders and a small codebase; the company grew to about 40 people, with the engineering teams building on that architecture.
Python Go Django Celery PostgreSQL Redis RabbitMQ Elasticsearch Pandas
Python Flask SQLAlchemy PostgreSQL
PHP MySQL JavaScript jQuery
First full-time job: an automated monthly invoicing system, from the calculation logic to the rendered documents.
C# Java JavaScript
Taught adults the basics of computers and Microsoft Office. The explaining-things career started before the engineering one.
Confetti, my own conference tracker: two Claude agents research conferences and return structured JSON, deterministic Python validates it and writes the data. Each agent gets two tools and a four-sentence brief and nothing else: no project context, no MCP servers, none of the coding-agent scaffolding. Every run carries a hard dollar ceiling and logs what it cost.
Either production problems I had to solve, written up with diagrams: rate limits, circuit breakers, spotting a bad deploy. Or observations about humans interacting with code: cognitive biases that make smart devs write silly code, how we lie to each other with stats, ...
My partner's business site, where I am the entire team: the full-stack engineer, the designer, the sysadmin, and even the marketing.
I started teaching before I learned programming. Lately, I speak at big developer conferences like Devoxx Belgium (3,000+ developers, sells out in minutes), J-Fall (the biggest Java conference in the Netherlands) and PyCon Italia, all in my own time, no sponsors and no employer agenda. Some talks come straight from the systems above: what happens when a third-party API goes down, and how to tell if a deploy broke production. The others are about the lies we tell ourselves: how we estimate, how we argue, ... The full record and the talk catalogue.
With a semester in 2008 at the University of Southern Denmark, in Odense, on exchange.
Helping teams make sense of complex systems: a staff engineering role, architecture work, untangling a codebase that's outgrown its team, or developer-facing work where explaining clearly is the job.
Reach me on LinkedIn.