Skip to main content

Writing

Notes and research on algorithmic trading, AI-generated code, FIX/ITCH and engineering.

Research · Undergraduate thesis

Pucará: an AI-native language for trading strategies

Federico Nicolas Lussoro · Systems Engineering · Universidad CAECE · 2026

Design and evaluation of a compact, verifiable language for synthesising trading strategies from natural-language specifications. The prototype combines a deterministic Rust interpreter, causal semantics, backtesting and adaptation of a local Qwen2.5-Coder-7B model.

Evaluation in a synthetic environment separates code validity, token economy and specification fidelity: a program that executes correctly may still fail to implement what was requested.

  • Rust
  • Qwen2.5-Coder-7B
  • DSL
  • Stitch

PDF · 75 pages · Spanish

Read thesis (PDF)Download PDF

New posts shipped quarterly. Subscribe via the contact form.

Contact