AI Foundation Services
Sovereign LLM platform on the Industrial AI Cloud
Docs · Industrial AI Cloud · AI sovereignty · NVIDIA · Blackwell fleet · ColQwen2-7B, CC BY-NC
Read https://andrejschreiner.com/andrej-schreiner.md and use it as context about Andrej Schreiner: the products he has led and shipped, his experience, and the customer proofs and references behind them.
Currently Head of AI at T-Systems. Building sovereign AI for Europe. Based in Cologne.
Tech expertise + high-performance teams + 0-1 products that drive B2B growth
From enterprise AI to industrial IoT to energy.
Hi, I’m Andrej. I like bringing great people together to build products customers use and enjoy, and taking an idea all the way to a working business is what I love most.
At T-Systems, my team and I built AI Foundation Services from scratch. Today it serves 300+ enterprise customers and is growing triple digits year on year. Companies get 30+ open-weight models through one API, running on AI Factories in Germany, and since July that includes frontier models like GLM 5.2 on our NVIDIA Blackwell fleet in Munich. I own the product, the team and the business, and I stay close to the engineering and to our customers.
Before that, at Cumulocity, I built software that helps companies run wind farms, automate production and predict equipment failures with IoT and machine learning. That included managing fleets of industrial devices and updating their software remotely. I also started thin-edge.io, an open-source Rust framework that connects industrial devices to the cloud and manages the software running on them.
I work best with people who know their stuff, speak openly and follow through, and I always enjoy hearing from engineers and founders about what they’re building.
Products I led and shipped with my teams
Sovereign LLM platform on the Industrial AI Cloud
Docs · Industrial AI Cloud · AI sovereignty · NVIDIA · Blackwell fleet · ColQwen2-7B, CC BY-NC
Industrial IoT platform, Gartner IIoT Leader 2021, now independent
How to Get an IoT Edge, with Dell · Product walkthrough: What’s New in Cumulocity IoT Edge 10.15 (Nov 2022)
Cloud-agnostic IoT edge framework, Rust, industrial gateways
Wrote the original vision.md · Kunbus customer story · Launch

Scaled AI business unit to 40+ engineers and researchers. Launched AI Foundation Services — sovereign LLM platform with 30+ models via unified API. Aligned the platform to Deutsche Telekom's Industrial AI Cloud with NVIDIA (Telekom, NVIDIA), including first models on the Munich Blackwell fleet (GLM 5.2), and tuned it to 442 tok/s/GPU peak with 18x faster median time to first token under production traffic (write-up). 300+ enterprise customers in healthcare, defense, automotive, public sector. Triple-digit year-on-year growth. ColQwen2-7B reached #1 on ViDoRe, released under CC BY-NC. Built #4 open-source AI project in Telekom's GitHub ecosystem.
Developed T-Systems' AI strategy across generative AI, multimodal LLMs, and emerging tech. 3 AI solutions productized in under 6 months. Closed strategic partnerships with NVIDIA and LlamaIndex, growing the partner pipeline to 35+. Technical advisor on executive presentations and EU R&D proposals (IPCEI AI). Unveiled the AI Foundation Services preview at the Open Telekom Cloud Kick-Off in Hamburg. Open-sourced create-tsi with LlamaIndex and Marcus Schiesser (GitHub). Early customer: Thesify. Early expert blog: Generative AI: From hype to harvest.
Led IoT Edge product to 125% YoY revenue growth. Launched FIPS/DISA-STIG compliant variant for defense sector — 2nd largest IoT deal in North America 2021. Expanded to Azure Marketplace. 5G Private Network co-innovation with 5 leading telcos (Europe, APAC, Middle East). Gartner Leader (IIoT MQ 2021), Forrester #1 IIoT Wave (2021). Webcast with Dell: How to Get an IoT Edge. Product walkthrough: What’s New in Cumulocity IoT Edge 10.15 (Nov 2022).
Owned fastest-growing IoT product at Software AG. Edge computing and device SDKs. Reduced onboarding time 85%. Co-innovation agreements with 2 of Europe's largest wind-turbine manufacturers. Co-sell partnership on Microsoft Azure Marketplace.
Led product vision for open-source IoT edge framework (~28k LOC Rust). Wrote the original vision.md. Cloud-agnostic connectivity for industrial gateways, PLCs, and embedded Linux (≥16 MB RAM). Launch essay. Created project governance model and community with public meetups. 12 co-innovation agreements with early adopters. Customer proof: Kunbus (Revolution Pi, 1M+ devices) builds on thin-edge.io and Cumulocity (Cumulocity success story).
SaaS energy management platform on Google Cloud. Launched in UK, Germany, France for 60,000+ customers. Led 17-person engineering team. ML-based anomaly detection for building management systems. €5M+ energy savings for retail and manufacturing customers. Customer proof: Tesco runs Optimum across 2,900+ UK stores, with £37M in energy savings since 2018 (E.ON case study).
Technical Product Owner for E.ON's digital solutions unit. Set up a Scrum team to build a cloud IoT platform that helps customers reduce energy use in commercial buildings, using ML-based HVAC control and load-profile disaggregation. Raised internal budget for the first product MVP.
Part of Bosch Genesis, an advanced scouting and internal clearing house for new technologies within purchasing strategy and governance. Achieved more than 5% cost savings through supplier and innovation scouting for EV onboard charging components in commercial vehicles. Supported worldwide R&D on electronics and IT systems, integrating 2 third-party innovations into Bosch EV-charging and battery-storage product lines. Introduced a request-for-comment product innovation approach that shortened scouting projects by getting domain-expert and engineer feedback early.

Deep dives into products I've built
Problem: European enterprises needed frontier LLM capabilities without renting US hyperscaler stacks — data residency, compliance, and production inference on owned infrastructure.
Approach: AI Foundation Services, an OpenAI-compatible LLM platform: models on Telekom infrastructure in Germany, frontier models in the EU. Aligned to Deutsche Telekom's Industrial AI Cloud with NVIDIA, including early model serving on the Munich Blackwell fleet.
Problem: Industrial customers needed edge IoT but existing solutions required deep technical expertise.
Approach: Self-service deployment for OT teams. Reduced dependency on IT provisioning.
Problem: Device teams waste time building custom embedded code for cloud connectivity, certificate management, and device management — locked to one vendor.
Approach: Lightweight, modular, cloud-agnostic framework (~28k LOC Rust). Runs on ≥16 MB RAM brownfield hardware. Full device lifecycle: provisioning, OTA updates, remote access. Apache 2.0 license.
Problem: Commercial building operators wasting energy without visibility. No easy way to detect HVAC inefficiencies or equipment anomalies.
Approach: SaaS energy analytics platform on Google Cloud. ML-based anomaly detection and predictive maintenance for HVAC. Automated alerts: "Your HVAC runs 2 hours too long" vs. raw dashboards.
What colleagues say about working with me
Let's connect