ralph balazs

Hello, I'm

Ralph Balazs

Software engineer going from low-level code to real-world machines

I build at the intersection of systems, applied machine learning and robotics. Grounded in C/C++ and Linux. Driven by what happens beyond the screen.

About Me

I like knowing how things work. Then making them work.

I’m Ralph, a software engineering student at École 42 Paris and a member of 42AI. I started with systems programming: memory, processes, drivers and the details that higher-level tools hide.

Today, I apply that foundation to machine learning and robotics. I care about tested code, clear trade-offs and being precise about what a project can actually do.

  • École 42 Paris
  • 42AI
  • Paris, France

Education

École 42 Paris — Software engineering, AI & Algorithms track. 2023 — present.

Teaching

Tutor at École 42 — C, algorithms and systems programming. March — September 2025.

42AI

Member of 42AI, the school’s artificial intelligence association.

Chess & community

Founded the school’s chess club, organising classes, tournaments and training sessions.

Skills & Technologies

What I use, grouped by how I use it.

Foundations

  • C
  • C++
  • Linux
  • x86 Assembly

Applied work

  • Python
  • scikit-learn
  • pandas
  • LeRobot
  • SmolVLA

Build & validate

  • Git
  • Docker
  • pytest
  • QEMU
  • K3s
  • Argo CD

Technologies I work with

  • C
  • C++
  • x86 Assembly
  • Linux
  • Python
  • scikit-learn
  • pandas
  • Matplotlib
  • LeRobot
  • SmolVLA
  • Git
  • Docker
  • pytest
  • QEMU
  • GRUB
  • K3s
  • Argo CD
  • Bash
  • KiCad

Featured Projects

From learning a physical task to writing a keyboard driver. A selection of projects, with the decisions and limitations behind them.

More to explore

Conceptual decision tree for a game bot: observations, candidate moves and evaluation OBSERVEMOVE AMOVE BMOVE CEVALUATEEVALUATEEVALUATE A decision process, not a game replay.

Competitive programming

Better decisions. One experiment at a time.

Competitive game bots built around search, opponent modelling and measured iteration. A place to test whether an idea actually makes a better decision.

  • C++17
  • Python
  • Experimentation
Scheduling data↓Leakage-safe features↓Cross-validation↓Model vs. human baseline
Pipeline architecture · no patient data

Machine learning

A prediction is only as good as its evaluation.

A surgery-duration pipeline benchmarked against the hospital’s existing estimates. Designed to catch data leakage and make its limits explicit.

  • Python
  • scikit-learn
  • pandas

Get In Touch

Paris, France · Open to a conversation

Let’s build something real.

Working on systems, applied ML or robotics? I’d love to hear what you’re building.

Get in touch balazsra@gmail.com

Let’s connect

I’m a software engineering student at École 42 Paris. Whether it’s about systems, applied machine learning or robotics, I’d love to talk.

Emailbalazsra@gmail.com
LocationParis, France

Follow me