Tareq Abu El Komboz
M.Sc. Computer Science Student / Incoming PhD Applicant / RL Researcher
I am a Master’s student at the University of Stuttgart (Focus: Intelligent Systems), graduating in March 2026. My research interests lie at the intersection of Deep Reinforcement Learning and Large Language Models. My journey so far has been supported by the Baden-Württemberg Scholarship during my time at the University of Adelaide and refined through industrial research at Porsche AG. I am actively seeking PhD positions for late 2026 focused on Machine Learning and Reinforcement Learning. My goal is to contribute to the top-tier research community (ICML, ICLR, NeurIPS) by developing robust, data-driven solutions for complex, high-dimensional systems.
Research Interests
- Learn-to-Optimize (L2O): Developing RL-based optimizers (PPO/REINFORCE) that generalize across varying objective functions and constraints.
- Foundation Models for Engineering: Adapting LLMs for specialized domain tasks like automated test-case generation and trace-file analysis.
- Anomaly Detection in Discrete Sequences: Scaling Transformer architectures (LogBERT, LogLLM) for ultra-high-volume system logs.
Current Research & Industry Impact
Currently, I am a Working Student at Porsche AG, where I bridge the gap between theoretical ML and automotive reliability.
Master’s Thesis - “Deep learning for anomaly detection in vehicle control units in trace files”
In collaboration with Porsche AG, I am benchmarking LogBERT vs. LogLLM on real-world vehicle datasets and open-source benchmarks (BGL, Thunderbird). My research demonstrates that scaling model parameters and leveraging self-attention mechanisms significantly improves F1-scores in anomaly detection by capturing long-range dependencies in complex, non-stationary log environments that traditional expert-led methods miss.
Bachelor’s Thesis - “Parameter-dependent self-learning optimization”
My work explored the reinforcement learning perspective of optimization algorithms, representing them as policies. The research demonstrated that learned policies can outperform state-of-the-art algorithms in convergence speed and final objective function values for high-dimensional setups, specifically addressing the challenge of “free” optimization parameters and explicit constraints.
Selected Honors & Projects
- AI Incubator Award: Won the “Business Model Award” at AI Incubator Batch #3 for an AI-Fintech prototype, proving the viability of ML in complex financial decision-making.
- Scholarships: Recipient of the Baden-Württemberg Scholarship for international academic excellence.
- Technical Writing: I actively synthesize complex papers (e.g., AlphaZero/AlphaGo) into technical breakdowns to foster community understanding of Reinforcement Learning. 👉 Read my blog post on Mastering the Game of Go