Skip to content
View comprido96's full-sized avatar
8 followers · 13 following

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Block or report comprido96

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
comprido96/README.md

Federico Gambassi

AI Engineer — LLM systems & production backend (Python · Rust) · Florence, Italy

I build production LLM systems with industrial evaluation rigor — frozen eval sets, regression gates, red-team suites — and the backends underneath them in Rust and Python.

  • 🏠 Founder of Dymora — a smart-building SaaS running live with paying customers: Rust/axum API (800+ tests), Stripe subscription billing, staged CI/CD, zero-touch device-fleet provisioning → engineering case study
  • 🤖 Day job: production LLM agents — planner/executor/synthesizer architectures, provider-agnostic tool calling (OpenAI + Anthropic), RAG, eval infrastructure larger than the source it tests
  • 🔬 Fine-tuning + eval methodology, end to end: frozen hand-verified eval set, prompted frontier baselines, LoRA on Apple Silicon beating gpt-5-mini (98.5% vs 90.25% exact-match, $0 cost) — with the ablation surprise and failure taxonomy published, not just the headline number → comando
  • ⚡ Solana/Rust: OTC swap protocol with Pyth oracle pricing and collateral-ratio enforcement → LetsDAT-Otc-Swap
  • 🎓 M.Sc. Mathematical Engineering (Politecnico di Torino) — thesis on model-based offline deep RL → Model-Based-Offline-Deep-RL

Highlights: HomeEdge-Orchestrator · comando · LetsDAT-Otc-Swap · TMDB-Agent · Model-Based-Offline-Deep-RL

📫 fedegambassi96@gmail.com · LinkedIn

Pinned Loading

  1. comando comando Public

    LoRA fine-tune of Qwen3-4B beats gpt-5-mini on Italian smart-home NLU — frozen evals, ablation, failure taxonomy

    Python

  2. dymora-engineering dymora-engineering Public

    Engineering case study: solo-building and operating a production SaaS with hardware in the field — Rust/axum, 806 tests, staged deploys, zero-touch device fleet

  3. HomeEdge-Orchestrator HomeEdge-Orchestrator Public

    HomeEdge Orchestrator — A Rust Control Plane and Agent Runtime for Smart-Home Node Management

    Rust

  4. LetsDAT-Otc-Swap LetsDAT-Otc-Swap Public

    Solana/Anchor OTC swap protocol — zBTC↔sBTC mint/burn with Pyth oracle pricing, collateral-ratio solvency checks, BPS fees; localnet/devnet integration tests

    Rust 1

  5. solana-reward-claim solana-reward-claim Public

    Anchor program: backend-authorized SPL reward claims with ed25519 instruction introspection, offset-pinning defense, and replay protection

    Rust

  6. Model-Based-Offline-Deep-RL Model-Based-Offline-Deep-RL Public

    Model-Based Offline Deep RL - experiments with the COMBO algorithm for oflline model-based RL. This work was done during my master research thesis, in which I had to optimize the energy consumption…

    Python 3