Barış Sözüdoğru
Engineering manager in Munich. Still building.
Now
At taico, I manage the infrastructure and architecture of a GCP platform serving major European enterprises in media analytics, retail, and streaming. Separately, I consult for Telefonica (O2) on 5G edge computing and computer vision for autonomous systems.
Day to day, I make large GCP platforms scale across cloud, data, and AI. My academic background is in robotics, computer vision, and autonomous systems.
In my own time I build side projects, from an artificial life simulation to memory tooling for coding assistants. Details on each below.
Work
Engineering Manager, taico GmbH
2024 - present
Leading architecture, infrastructure, and security for a GCP-based analytics platform serving enterprise clients in media analytics, retail, and streaming.
Consultant, Telefonica (O2) Germany
2024 - present
Advisory on 5G, edge computing, and computer vision at the Technology Strategy & Innovation department.
Founding Engineer, taico GmbH
2021 - 2024
Built the platform from scratch: data pipelines, APIs, cloud infrastructure. Helped take the product from zero to funded, with the first enterprise clients.
Education
Technical University of Munich.
Technical University of Munich.
Research
Master's thesis with Telefonica. Compared 5G and Wi-Fi for autonomous systems on an industrial forklift. Presented to Telefonica supervisory board.
Scene-aware and social-aware neural networks for vehicle behavior prediction. Chair of Cyber Physical Systems, TUM.
Large-scale ground truth dataset for automated structured radiology reports. Chair of Computer Aided Medical Procedures, TUM.
Localization for autonomous driving via GNN-based point cloud registration. Chair of Robotics, AI and Real-time Systems, TUM.
Bachelor's thesis. Deep learning for image classification on Raspberry Pi. Chair of Data Processing, TUM.
Projects
Petri Labs
Artificial life simulation where neural network organisms evolve, hunt, and speciate in real-time in the browser.
Cortex AI
Persistent memory for coding assistants, running locally on macOS. Carries context between sessions, learns how the user works, and reports calibrated confidence for what it retrieves.
ADK Agent Playground
Seven multi-agent systems built on Google ADK, with shared components for observability, security, and memory, plus an evaluation round that measures which parts of an agent setup actually earn their place.
Deep Research Digest
Federated academic search across arXiv, PubMed, and Semantic Scholar.