Math research focus
Homogeneous dynamics and representation theory.
PhD student in mathematics
Doctoral student at the University of Zurich, conducting research in homogeneous dynamics, representation theory, and trace formulae under the supervision of Alex Gorodnik.
Math research focus
Homogeneous dynamics and representation theory.
AI systems & infrastructure
Engineering RAG systems, LLM evaluation harnesses, AWS-based cloud infrastructure, data pipelines, and reproducible research workflows.
Supervisor: Prof. Dr. Alexander Gorodnik.
French M2 program in Langlands program and representation theory. Full scholarship supported by the Archimedes Institute.
Received an Individual Sports Scholarship in Dec 2018.
Worked on US options selling strategy research, focusing on entry filters, premium/risk constraints, sizing rules, exit logic, and replay-based validation.
Open Source Contributions
Contributed a merged fix for global runtime-source integrity: manifest-verified, single-provider resolution prevents command and agent content from being mixed across sources. The update path stages artifacts before publishing state, preserving the prior runtime when validation or writing fails.
Self-Directed Learning
Technologies, frameworks, and systems explored through independent study, certifications, and applied projects.
Interesting Side Projects
A controlled research showcase testing whether direct Lie-bracket supervision helps visual dynamics models recover order-dependent action effects. The preliminary one-seed result improved the local swapped-order measure, while exposing an unresolved gap in longer-horizon composition.
An independent, dependency-light feasibility experiment maintained inside my RISC-V UnifiedDB fork. It uses canonical environment receipts and offline-bundle replay to keep the study reproducible without the full upstream toolchain.
A Chinese-first German survival-vocabulary MVP built around practical scenes such as the doctor and supermarket, with lightweight spaced review and persistent local progress.
A local CLI and Codex skill that turns job descriptions and recruiter notes into consistent Obsidian records, including application status and role-track inference.
A local-Ollama automation for turning daily notes into a monthly review, with a small operational wrapper for checking and starting the local model service.
A compliance-first, read-only pipeline for turning a market topic into research plans, structured evidence, market signals, and traceable reports.
A bilingual, link-verified learning map for ML theory, PyTorch practice, LLMs, and AI system-design interview preparation.
Beyond Work
Fields, courts, water, and trails.
Strategy, long arcs, and imagined worlds.
Home cooking and everyday experiments.