Thomas Wölkhart

ML & AI Engineer — SAP

Walldorf / Munich · Esports (Team Liquid)

• Converted from working student to full-time after completing M.Sc. studies.
• End-to-end ownership: ML model development, full-stack tooling, and production deployments for League of Legends esports analytics.

ML & AI Engineer (Working Student) — SAP

Walldorf / Munich · Esports (Team Liquid)

• Returned after HMS research stay; continued building and deploying ML systems for the Team Liquid partnership.

Visiting Graduate Student — HMS / Brigham & Women’s Hospital

Boston, MA · Surgical Planning Lab (Radiology)

• Research on quantitative analysis of Digital Subtraction Angiography (GAE).
• Tooling: Python, PyTorch, nnU-Net, OpenCV, scikit-image, NIfTI.

ML & AI Engineer (Working Student) — SAP

Walldorf / Munich · Esports (Team Liquid)

• Converted from internship mid-January; built, deployed, and operated ML pipelines in HANA Cloud for League of Legends analytics.
• Automated Team Fight Detection, saving >1000 hours/year.

ML & AI Engineer (Intern) — SAP

Walldorf / Munich · Esports (Team Liquid)

• Joined the esports analytics team; converted to working student role mid-January.

AI Engineer (Working Student) — Mercedes-Benz AG

Munich · Data & AI (PoCs)

• Time-series models for sales forecasting & scenario simulation.
• RAG prototype to automate market research workflows.

AI Engineer (Research Assistant) — TUM School of Management

Munich · pAIge (LLM Student Support)

• Backend for pAIge Chatbot; reduced student-support workload.
• RAG pipelines on Azure with vector search and feedback loop.

Generative AI (Working Student) — appliedAI

Munich · GenAI Prototyping

• Prototyped and deployed GenAI apps (RAG, Voice-AI) for internal & client use.

Member · Team Lead · Advisor — TUM.ai Student Initiative

Munich · Student Initiative (Operations & Projects)

• Led >10 members; managed six-digit budget and industry AI projects (Infineon, UniCredit, ProSiebenSat.1).

Business Development & Data Automation (Intern) — LOEWI (acq. by bioniq)

Munich · Growth & Data Automation

• Python automation for B2B lead generation; micronutrient bioavailability database.

M.Sc. Sport & Exercise Science — Technical University of Munich (TUM)

Grade: 1.0 · Thesis grade: 1.0 · Sport Informatics, Neuromechanics & Physiology

Visiting Graduate Student — Harvard Medical School / Brigham & Women’s Hospital

Surgical Planning Lab · Medical Imaging / Computer Vision

B.Sc. Sports Science — Technical University of Munich (TUM)

Grade: 1.3 · MuSkLE Summer School (France, 2023)

Secondary School (A-levels) — Bundesrealgymnasium Köflach (AT)

Grade: 1.0 · Focus: IT, Biology, Chemistry, Physics

Master’s Thesis — Towards Continuous ADL Detection from Non-Intrusive Load Monitoring: Scalable Machine Learning Approaches for Health-Tech Applications

Grade: 1.0 · Veli

• 1D-CNNs for device-level activity recognition from high-frequency time-series power data.
• Synthetic data pipelines and a Bayesian hierarchical model for scalable, continuous ADL detection.

Wölkhart et al. — Intraoperative Quantification of Hypervascularity during Genicular Artery Embolization (SPIE Medical Imaging 2026)

• Computer-vision pipeline for DSA; automatic isolation of neovascular blush; standardized intraoperative metric.
• Inference code & model weights: GitHub

1st Place — TUM Football Analytics Hackathon

• Derived a fatigue score from 3D skeletal tracking data.

1st Place — MBZUAI Hackathon (Abu Dhabi)

• Realtime voice-driven website editor [post].

TUM × Data Reply — Data Innovation Lab

• Multi-agent system on AWS; evaluated LangGraph, AutoGen, MetaGPT [write-up].

1st Place — Alexandria ML-Agents Hackathon

• Shared-memory architecture across LLM apps; backend + frontend implementation.

2nd Place — BKW Engineering Hackathon

• Platform for room classification and adaptive cost computation.

3rd Place — TUM.ai Makeathon (Mercedes-Benz & Salesforce)

• Conversational Sales Bot prototype [devpost].

1st Place — TUM & Microsoft LLM Hackathon

• RAG chatbot for student support; automated email responses; vector DB ingestion [code].

Deutschlandstipendium — Scholarship