AI engineer, systems architect and gardener
Felix Lorenz
I build AI systems that run real businesses, and the platforms that keep them safe, observable and fast.
Senior engineer in Berlin. My path runs from physics competitions and web security, through distributed-systems research and big-data platforms, to autonomous AI agents in production.
- Focus
- AI agents, generative media, platform ops
- Based
- Berlin, working remote
- Speaks
- German, English, French, some Arabic
Companies and labs I've built with
The arc
From physics prizes to production agents.
- 2008 Science Jugend forscht and ICYS awards in physics
- 2008 Hacking Own web studio, security analysis of web apps
- 2014 Infrastructure Bare-metal clusters and infrastructure as code
- 2018 Research Runtime verification, IoT security, 7 papers
- 2019 Big data Kafka and Flink pipelines for city water grids
- 2022 Cloud & SRE Kubernetes platforms and SLOs for large enterprises
- 2023 AI platforms GPU orchestration, fine-tuning, secure AI
- 2025 Agents Generative media and autonomous agents in production
01
About
I'm Felix, a software engineer who builds AI systems that run in production. I care about the whole thing: the agent or model, the platform under it, and the guardrails, monitoring and cost control that let a business trust it.
I came to this through science and hacking. As a teenager I won physics prizes at Jugend forscht and the International Conference of Young Scientists, and ran a small web studio that also did security analysis of web applications. I studied media and communication, then computer science at TU Berlin, and worked on neural networks, formal verification and recommender systems along the way.
Eight years in applied and university research, first running compute infrastructure and then as a researcher, taught me distributed systems: runtime verification for autonomous robots, sensor fingerprinting for IoT security, and Kafka and Flink pipelines for monitoring city water networks, published at IEEE Big Data and other venues. After that I took platforms to production as a DevOps and reliability engineer in the automotive and public sectors.
Since 2023 I have worked at AI-first startups: GPU orchestration and fine-tuning, secure AI for the medical and public sector, multimodal generation pipelines, and now autonomous agents. I like owning things end to end, from architecture to on-call.
When I'm away from the keyboard I'm usually in the garden, on a bike, or dancing.
- German
- Native
- English
- Fluent
- French
- Conversational
- Arabic
- Basic
02
Experience
Where I've worked, most recent first.
-
Founding Engineer, AI & Ops
Agent-first startup in generative media.
- Design and build autonomous AI agents that run business processes end to end
- Safety architecture for agents that act on their own: guardrails, limits, kill switches, testing before every release, audited access
- Own infrastructure and operations: deployments, observability, data pipelines, payments and app releases
-
AI Engineer, Full-Stack
Generative platform that turns prompts or full scripts into illustrated, animated and narrated stories.
- Orchestrated multimodal pipelines (text, image, video, speech) across several providers, with per-story model switching, fallbacks and provider monitoring
- Designed a consistency layer that keeps characters and locations visually stable across a story, including countermeasures against identity drift
- Shipped new state-of-the-art models to production, sometimes on their release day (Nano Banana 2, Seedance 2.0, Veo 3.1, Gemini 3)
-
AI Engineer
AI platform for regulated sectors.
- Built, ran and scaled an application that transcribes and analyses professional conversations in the medical and public sector
- Implemented strict security and data protection requirements: OAuth/OIDC, API gateway, secrets management
- Integrated with the platform's microservices over Apache Kafka; monitoring and alerting with Prometheus, Grafana and SigNoz
-
Full-Stack Engineer, DevOps Lead
AI audio startup.
- Led the team technically: system architecture, DevOps stack on GCP, monitoring and dynamic scaling
- Built a platform that orchestrates data-intensive workloads across heterogeneous CPU/GPU pools
- Core developer of an open source framework for AI audio workflows; fine-tuned Llama 3, MusicGen and Whisper
-
Site Reliability Engineer
Public-sector security technology provider.
- Guided product teams across the company from SLA, SLO and SLI definition to go-live with alerts, incident playbooks and post-mortems
- Designed service-level metrics for production platforms on Kubernetes and OpenShift
-
DevOps Engineer
Automotive: platform for processing and visualising large vehicle sensor datasets.
- Migrated the platform from Docker hosts to Kubernetes and set up continuous delivery in Azure DevOps
- Built the authentication and authorisation service (FastAPI, enterprise SSO, SAML, Keycloak)
- Rebuilt the local dev environment with K3s and devcontainers for cloud-like development
-
Founder, Full-Stack Engineer
Community platform for healthy daily practice.
- Designed and built an open source web platform for organising self-help groups and practice communities
- Built the whole DevOps stack for development, delivery and monitoring of the platform
-
Research Associate, Cloud Engineer
University research group for distributed systems.
- Architected a scalable, fault-tolerant cloud platform for monitoring municipal water networks (federally funded research project)
- Built highly scalable analytics pipelines in Apache Flink; verified QoS guarantees with chaos testing and log analysis
- Prototyped secret-free authentication for IoT devices from fingerprints of their analog sensors
-
Research Associate
Applied research institute, software quality.
- Lead researcher in a federally funded project: a runtime verification tool for collaborative embedded systems
- Built a message bridge between real and simulated CAN bus systems
- Studied machine learning for autonomous driving under safety certification requirements
-
System Administrator
Applied research institute, scientific computing infrastructure.
- Ran bare-metal compute infrastructure and networks; automated provisioning as infrastructure as code
- Designed a highly available backend cluster for an open data portal
Earlier
- 2 yearsStudent Research Assistant, Mannheim. Transport research, data processing and automation.
- 2 yearsFounder, Lörrach. Web and IT projects for companies, security analysis of web applications.
03
Research & education
7 peer-reviewed papers on big data, IoT security, verification and recommender systems.
Publications
- 2021 Gral: Localization of floating wireless sensors in pipe networksIEEE International Conference on Cloud Engineering (IC2E)
- 2020 A scalable and dependable data analytics platform for water infrastructure monitoringIEEE International Conference on Big Data
- 2020 Fingerprinting Analog IoT Sensors for Secret-Free AuthenticationIEEE IoTSPT-ML Workshop
- 2019 Effectively Testing System Configurations of Critical IoT Analytics PipelinesIEEE International Conference on Big Data
- 2019 Towards runtime verification of collaborative embedded systemsSICS Software-Intensive Cyber-Physical Systems, Springer
- 2018 Online-monitoring autonomous transport robots with an R-valued temporal logicIEEE International Conference on Automation Science and Engineering (CASE)
- 2017 Countering contextual bias in TV watching behavior: introducing social trend as external contextual factor in TV recommendersACM TVX
Education
- 2016 - 2018 M.Sc. Computer Science Technische Universität Berlin Thesis: Online Monitoring of Collaborating Autonomous Transport Robots
- 2013 - 2016 B.Sc. Computer Science Technische Universität Berlin Thesis: An LSTM Recurrent Neural Network for 9x9 Go
- 2010 - 2013 B.A. Media and Communication Studies University of Mannheim Thesis: Using natural language processing to analyse political discussions on blogs
Awards
- 2009First place, Innovationspreis Physik
- 2009Third place, Jugend forscht state competition
- 2008Second place, International Conference of Young Scientists (ICYS)
- 2008First place, Jugend forscht regional competition
04
Skills
What I reach for, from agents down to the metal.
AI & agents
Models & media
Machine learning
Languages
Backend & apps
Data & streaming
Cloud & ops
Reliability & security
05
Side projects
Things I grow after hours: tools, games and experiments, most of them live on this server.
-
NawsMapper
2026Open source projection mapping for designers: draw surfaces over real objects, fill them with shaders, images or video, and animate them with Photoshop-style controls.
Source -
Ogbo
2023Ogbo is a virtual monastery that offers daily activities for a relaxed & healthy life.
-
FamTown
2025FamTown is a family home dashboard. It helps households organize shared calendars, files, notes, photos, finances, and has a collaborative townsquare that can be decorated.
Source -
Sh8r
2026sh8r is a live VideoJockey tool for submitting and performing GLSL shaders.
Source -
I Ching
2024Digital oracle based on the ancient Book of Changes
-
Where's Alan?
2026A pixel-art party game of paranoia: a handful of humans and a hidden AI chat, doodle and whisper, then vote on who to unplug.
-
The Space of Change
2026The 64 hexagrams as one living body: a six-dimensional cube that morphs through Shao Yong's circle, King Wen's sequence and the breath of the year.
-
Kyoki
2024Simple two-player online card game based on the original game 'Coup de Grace'.
Source -
Depths of Change
2026A pixel-art dungeon crawl where you are a hexagram: change your lines to fight, open sealed chests and pass gates that only open to the hexagram they name.
-
Yao Rush
2026A neon arcade rush through yin and yang gates that teaches the eight trigrams and all 64 hexagrams of the I Ching.
-
Daily Oracle
2026A daily reading that weaves the real sky, the Chinese almanac day and a cast I Ching hexagram into one.
-
Radio NOW
2026A small radio station in a pagoda on a mountain lake, where AI-revived thinkers, poets and comedians muse on the world, laugh at the news and invite you back into the present moment.
Source -
Dismount Derby
2026A 3D physics game about getting down a seeded procedural mountain as fast as possible, most runs ending as a ragdoll.
Off the clock
And the actual garden.
06 · Contact
Let's build something, together.
Open to conversations about AI engineering, agent systems, platform architecture and founding-team roles.
felix@breathdance.net

