Tomás Erdmannsdörffer
builds ML systems where
physics meets code.
Machine learning engineer with a foundation in scientific computing and deep learning. I work the full stack — training and compressing models, building LLM pipelines, shipping backends to production. Currently researching cardiac fiber modeling with Physics-Informed Neural Networks at the University of Graz.
Currently
may 2026Researching cardiac fiber mechanics with PINNs.
I'm three months into a visiting research stay at the University of Graz, working with Dr. Federica Caforio on physics-informed neural networks for myocardial tissue modeling. Most of my time goes into my MSc thesis project: applying PINNs to cardiac modeling.
On the side, I'm building browser-side ML demos — the two on this page — and competing in OpenAI's Parameter Golf, a sub-16MB language model competition.
- research WarpPINN-Fiber adaption — PINNs for cardiac modeling.
- writing Compression thesis manuscript — knowledge distillation + pruning + quantization on PINNs.
- competition OpenAI Parameter Golf — sub-16MB language model on FineWeb bpb, RTX 5060 Blackwell.
- building Browser ML demos — PINN playground & RAG over scientific papers (both live on this page).
Live Demos
both run in your browser · no installPINN Playground
Train a physics-informed neural network in the browser to solve the 1D viscous Burgers equation. No labeled data — just the PDE and autodiff.
Ask My Research
RAG system over a curated corpus of 24 scientific ML papers — PINNs, neural operators, compression, LLM inference. Retrieval runs in your browser.
FNO vs. Solver
Watch a neural operator solve Burgers ~10-50× faster than a finite-difference solver — same answer, fundamentally different algorithm.
My Journey
2020 → todayVisiting Researcher research
Working on PINNs for cardiac modeling in JAX.
MSc in Engineering Sciences education
Researching PINNs for cardiac fiber modeling applied to PDEs — the foundation of my work in Graz. The MSc deepens what started as a thesis on compressing PINNs.
DevOps Intern internship
Maintained and optimized 3 legacy applications, improved the production CI/CD pipeline to reduce operational complexity and deployment time, and collaborated on code migration via Bitbucket.
Undergrad Thesis — PINN Compression research
Applied knowledge distillation, pruning, and quantization to PINN surrogates for non-Newtonian (Carbopol) fluid simulation. Built the baseline PINN end-to-end, then quantified precision vs. efficiency trade-offs across techniques.
Teaching Assistant (×6 courses) teaching
Supported ~30 students per semester across Operating Systems, Low-Level Programming (C/C++), Mobile Apps, Web Technologies, Automata & Computability, and Programming Fundamentals.
AI Research Intern internship
Researched generative AI tools for retail. Built a functional prototype for automated product descriptions, transforming a manual process into an automated flow. Delivered a technical proposal that continued in development post-internship.
BSc Civil Engineering in Computer Science education
Coursework: Low-Level Programming, AI, LLM, Computer Vision, Algorithms & Competitive Programming, Databases, Web Technologies, Data Structures.
Research & Software
beyond the demosMSc Thesis MSc thesis
Working on PINNs for cardiac modeling in JAX.
Parameter Golf competition
OpenAI public competition: build a language model under 16MB, scored by bits-per-byte on FineWeb. Designing custom tokenizers and quantized Transformer architectures.
- Targeting RTX 5060 Blackwell (sm_120, PyTorch cu128 nightly).
- Background in PINN compression directly applicable.
Turbo Files production
Fault-tolerant data transfer system shipped to a mining company in production. Fragments large files for resilience over intermittent connectivity.
- Google Drive + AWS S3 integration with variable file sizes.
- Built end-to-end in Django.
Toma de Ramos leadership
Led a team of 7 building a web-based academic scheduler. Used by the entire Universidad de los Andes for one full year.
- Credit validation + schedule conflict detection.
- Django backend, React frontend.
Stack
ML / AI [core]
- PyTorch
- JAX
- TensorFlow · Keras
- HuggingFace
- CUDA
- PINNs · diff. sim
- RAG · LLM agents
Languages [code]
- Python
- JavaScript
- C / C++
- C#
- SQL
- LaTeX
Backend & Web [apps]
- Django
- Node.js
- React
- Ruby on Rails
- REST APIs
DevOps & Cloud [ops]
- Docker · Git
- AWS (S3, Lambda)
- Jenkins · Bitbucket
- CI/CD
- Linux · WSL2
Contact
Let's build something that actually runs.
Open to ML/AI engineer roles — remote, hybrid, or on-site globally. Available part-time now; full-time after MSc completion.