Hello, world

Michalis Marcoux

Systems engineer. Rust, ML, and the space in between.

I write software that sits close to the metal and thinks in matrices. Currently building financial validation systems by day and training neural networks by night.

My work lives at the intersection of Rust systems programming and Python ML pipelines — places where performance and correctness aren't optional. I care about code that's honest: no hidden allocations, no silent failures, no magic.

When I'm not debugging distributed systems at 2 AM, I'm probably reading about compiler theory or working on my next side project.

Rust Primary Language
ML PyTorch & Keras
Fin Validation Systems
Sys Low-Level Design

Selected Work

DPM-XL Validator

Financial validation engine built in Rust. Evaluates regulatory reporting rules against large datasets with sub-millisecond latency. Custom macro DSL for declarative rule definitions.

  • Rust
  • Proc Macros
  • MSSQL
  • Financial Data

Market Prophet

LSTM + CNN ensemble for stock market prediction. Custom data pipeline with Polars, hyperparameter sweep with Optuna, deployed as a REST API.

  • Python
  • PyTorch
  • LSTM
  • Polars

Static Forge

A static site generator with markdown parsing, templating, and syntax highlighting. Built from scratch — no frameworks.

  • Rust
  • Markdown
  • Templating

Zero-Access Vault

Encryption library implementing zero-access patterns with Argon2id key derivation. The server never sees plaintext.

  • Rust
  • Cryptography
  • Argon2id

Imbalance Lab

Data mining toolkit for imbalanced classification. Implements SMOTE variants, cost-sensitive learning, and threshold optimization.

  • Python
  • Scikit-learn
  • Data Mining

Posts:

Let's talk.

Whether it's about Rust, ML architectures, or an interesting problem you're working on — I'm always up for a conversation.