Tobias Fuchs

Tobias Fuchs

Software Engineer & PhD in Computer Science
Karlsruhe, Germany • LinkedIn • Google Scholar • GitHub

PhD in Computer Science specialized in robust AI. Proven track record in high-performance software engineering at AWS and SAP, complemented by research on reliable ML methodologies for noisy data conditions with publications at TMLR, UAI, and ECML. Expert in bridging the gap between theoretical ML and production-grade systems.

Work Experience

Doctoral Researcher Jul 2023 - Jun 2026
at Karlsruhe Institute of Technology • Karlsruhe, Germany
  • Developed and optimized robust ML pipelines for real-world datasets, implementing principled statistical methods and custom data-cleaning algorithms to mitigate the impact of noisy, incomplete, or ambiguous annotations.
  • Open-sourced and maintained a modular research codebase on GitHub, prioritizing reproducibility and clean API design to facilitate further research in weakly supervised learning.
  • Supervised and mentored 5+ students through the lifecycle of their thesis and seminar projects, providing hands-on technical guidance in algorithm implementation, system benchmarking, and rigorous experimental evaluation.
Software Engineering Working Student Apr 2022 - Sep 2022
at SAP • Walldorf, Germany
  • Developed distributed system features for SAP HANA, a high-performance in-memory database, specifically focusing on the reliability of communication across large-scale clusters.
  • Implemented and shipped a topology consistency check within the database control plane to preemptively diagnose and resolve misconfigurations between distributed components like nameservers and storage nodes.
Software Engineering Internship Jun 2021 - Oct 2021
(with Full-time Return Offer) at AWS • Berlin, Germany
  • Developed automated optimization heuristics for the Redshift Advisor, leveraging metadata from petabyte-scale distributed clusters to generate actionable recommendations for database schema and query performance tuning.
  • Engineered and shipped a specialized internal system table to the Redshift production environment, enabling detailed telemetry on SQL function usage to facilitate the migration from generic string types to specialized column datatypes.
Software Engineering Working Student Jul 2017 - Jul 2020
at ITscope GmbH • Karlsruhe, Germany
  • Developed and maintained full-stack features for the ITscope platform using Java with Spring.
  • Decomposed a monolithic web service by migrating PDF business document generation into a standalone microservice, refactoring internal interfaces and designing APIs to improve resource isolation and memory management.

Education

Karlsruhe Institute of Technology Jul 2023 - Feb 2026
PhD in Computer Science (Dr. rer. nat.) • Karlsruhe, Germany
  • Completed my doctoral studies at one of Germany's best universities in less than three years.
  • Defended dissertation: "Accurate and robust weakly supervised learning with candidate labels" (DOI:10.5445/IR/1000190012), developing novel methodologies for training ML models under ambiguous real-world data conditions.
Karlsruhe Institute of Technology Aug 2020 - May 2023
Master of Science (MSc), Computer Science
  • Achieved the best grade for my Master's thesis on hyperparameter optimization for propositional satisfiability solving.
  • Specialized in algorithm engineering, implementing memory-efficient, dynamically-sized bit-vectors using red-black trees to achieve high-performance data manipulation with logarithmic complexity (GitHub).
Karlsruhe Institute of Technology Oct 2017 - Aug 2020
Bachelor of Science (BSc), Computer Science
  • Achieved the best grade for my Bachelor's thesis in the field of hypergraph partitioning.
  • Successfully passed courses on programming and algorithms, while still in high school, prior to formal enrollment.
Chalmers University of Technology Aug 2020 - Jun 2021
Erasmus exchange, Computer Science • Gothenburg, Sweden
  • Completed a one-year exchange program at Chalmers, focusing on parallel computing and distributed systems.
  • Implemented the Raft consensus protocol to ensure data consistency in decentralized environments.

Skills

Systems, Databases & Infrastructure: Theory (Raft/Paxos, CAP, ACID), SAP HANA, AWS Redshift, PostgreSQL, Redis, gRPC, Protocol Buffers, AWS (S3, EC2, Redshift, Lambda, DynamoDB), Docker, CI/CD, Ansible.
Programming Languages: Proficient in C++, Python, Java, SQL. Familiar with Rust, C#, Bash, Typescript, Javascript.
Machine Learning: Convex & Non-Convex Optimization, Statistical Learning Theory, Conformal Prediction, PyTorch, JAX, Transformers, Evidential Deep Learning, Amortized Variational Inference, Diffusion Models, Scikit-learn, XGBoost, LightGBM, CatBoost.
Spoken Languages: German (native), English (C1), French (B1).

Selected Publications

Partial-Label Learning with Conformal Candidate Cleaning Uncertainty in Artificial Intelligence (UAI)
Created a novel framework leveraging conformal prediction to prune ambiguous candidate sets without labeled validation data. [Paper]
Partial-Label Learning with a Reject Option Transactions on Machine Learning Research (TMLR)
Engineered a nearest-neighbor-based algorithm with a reject option to optimize predictive coverage and accuracy. [Paper]
Robust Partial-Label Learning by Leveraging Class Activation Values Machine Learning (ML)
Introduced a novel PLL framework based on subjective logic for explicit uncertainty quantification and resilience against adversarial perturbations. [Paper]

Selected Projects

nanoSAT — Minimal CDCL satisfiability solver GitHub
  • Engineered a lightweight CDCL SAT solver, implementing core features like unit propagation and conflict analysis.
  • Prioritized high readability and clarity by intentionally avoiding custom memory allocators.
Kreuzwort++ — Automated crossword generation Steam app
  • Developed a crossword puzzle app using TypeScript and Electron, reaching $1,000+ in profit.
  • Engineered a custom satisfiability encoding engine to automate grid generation.
Advanced Data Structures Implementation GitHub
  • Developed a suite of high-performance data structures, including a dynamically-sized bit-vector using red-black trees.
  • Focused on logarithmic runtime complexity and memory efficiency.

Achievements

Finalist in Germany's National Computer Science Competition (BWINF) Oct 2017
Participated in the final of the BWINF as one of the top 30 participants out of 1,700+ nationwide.
Award Winner, EnBW hackathon Nov 2019
Achieved a third place (500€ prize) in a hackathon about smart energy grids by building an accurate non-intrusive load monitoring and forecasting algorithm.