DATA • FINANCE • TECH

Mario Ciavarella

Finance and beyond explorations — tools I build to understand the problems they solve.

My collection brings together experiments on financial data, simple governance models and a few side projects I use to explore new ideas.

Working notes

  • Building things is how assumptions get exposed.
  • Most ideas fail not in code, but in framing.
  • Early optimization hides weak reasoning.
  • Refinement only matters after usefulness is proven.

Personal projects

Small experiments I keep returning to.

Each one documents the thinking so anyone can poke at it later.

AES.js preview

Dependency-free implementation of the Advanced Encryption Standard (AES) in modern JavaScript, covering the block cipher and the supporting primitives in plain code. The aim is a transparent, self-contained implementation, prioritizing code readability and portability for verifying logic or running in environments lacking native crypto APIs.

The library implements the full AES spectrum (128/192/256-bit) along with every major block mode, including GCM for integrity protection. It comes as a comprehensive toolkit with hashing and key derivation utilities, keeping the API compact but capable enough for prototypes and teaching environments.

v0.9.0 95 Jun 2025

Convergent Flow Protocol

Python ZK Proofs Cryptography Blockchain
Convergent Flow Protocol preview

A research blockchain prototype exploring DAG-based sequencing, ZK provers, intent auctions and UTXO state models. It implements an intent-centric execution layer where solvers compete via verifiable sealed-bid auctions with Groth16 proofs — all stake-backed and slashable for accountability.

The project combines Poseidon-based Merkle trees, commit-reveal auction mechanics, Circom/snarkjs circuits for proof generation and a SQLite persistence layer, serving as a lab for learning how accountability, deterministic ordering and zero-knowledge verification can be engineered as system properties rather than afterthoughts.

v0.1.0 60 Oct 2025

AI Audit Framework

Governance Privacy Merkle proofs Python
AI Audit Framework preview

A technical audit kernel for financial workflows, built to transform transparency from a regulatory concept into a data structure. The framework intercepts every AI decision—applying deterministic checks (LTV, DSR), monitoring statistical drift and protecting sensitive data (ε-DP)—before sealing everything in a tamper-proof Merkle ledger.

It is a minimalist, pure-Python project designed as an architectural sandbox: it demonstrates how to build systems where integrity isn’t an afterthought, but the foundation of the runtime itself.

v0.7.0 130 Dec 2025

Discord Poker Bot

Discord.js MySQL Multiplayer Real-time

A poker bot built with Discord.js, a MySQL-backed API and a lightweight control panel to run casual Texas Hold’em and Blackjack games inside private servers. It automates lobbies, seating, betting rounds, card reveals and bankroll tracking while persisting player stats and audit logs in a relational database.

First assembled as a weekend project and revived later, it now acts as a playground for asynchronous flows, fairness checks and incentives like streaks and leaderboards. The newer iterations focus as much on the surrounding tooling such as rate-limited endpoints, structured logging, access controls and a web dashboard, so the bot behaves like a small, observable multiplayer system rather than a one-off script.

v0.1.0 52 Sep 2024

Market Heatmap

Market data Simulation JavaScript Real-time data
Market Heatmap preview

A modular heatmap dashboard for equity markets, built as a browser client that flips between an offline simulation and a Finnhub feed. State, transport and rendering stay isolated: tiles live in a state manager, updates travel through an event-driven controller and the renderer batches DOM work on animation frames to keep interactions fluid.

It doubles as a UI-heavy JavaScript lab where I postpone picking a framework. Views own the DOM quirks (filters, sliders, modals) while a small services layer handles WebSocket traffic, rate limits, API keys and mode switches, leaving room to tune performance and reuse patterns in tougher dashboards.

v0.4.2 64 Dec 2025

mariociavarella.it

Astro TypeScript MySQL Web Design
Read More
Welcome to mariociavarella.it v0.4.2
Type 'help' for available commands.
System online.
~/site $

A digital garden built to be owned, not just rented. It abandons static site defaults for a bespoke Astro architecture featuring a persistent MySQL interaction layer, live synchronization with GitHub releases and an idle-state eco-mode. No templates, no heaviness.

Ideally, a portfolio is static; practically, a learning space needs state. This site bridges that gap, serving as a living proof-of-concept for how I architect software: strictly typed, component-driven and obsessed with performance.

???

</> coming soon

Request Access

This repository is private. Please request access below.

By submitting this form, you agree that I will use your details to reply to your request. See Privacy Policy.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Articles

Writing to understand what I’m building.

A curated set of notes and essays, available in full here.

Oct 7, 2025 7 min
89 5 1

Verifiable Intelligence: a Technical Framework for Auditable AI Governance

The race in artificial intelligence is no longer defined by scale or accuracy, but by the transparency of decision-making processes. As models enter regulated environments, performance becomes secondary to operational verifiability: systems must demonstrate why they work and how they fail.

Anthropic’s behavioural audits — stretching from the Claude 4 System Card to agentic misalignment studies — set a precedent: thousands of adversarial probes that surface hidden autonomy, reward hacking and shutdown resistance. The point is not to rubber-stamp safety but to map the surface of risk so each deviation is observable, classifiable and reproducible.

Finance already treats risk as an auditable quantity. Balance sheets are verifiable, processes traceable and metrics reproducible. Applied to AI, the same discipline demands architectures that emit deterministic logs, explicit constraints, measurable privacy budgets and systematic drift detection — so that every decision carries verifiable evidence of its own behaviour.

Oct 25, 2024 6 min
67 4 0

AES Without the Magic: A Practical Primer

If you read “encrypted at rest” or “secure channel” in a spec, AES is usually the thing actually moving the bits. It hides under labels like disk encryption, token stores, TLS offload and “secure cookies”, quietly encrypting data even when you never call it directly.

What hooked me was realising how unmagical AES is once you inspect it. On paper it’s a block cipher over finite fields; in a debugger it’s a 4×4 grid of bytes pushed through table lookups, rotations, XORs and round keys with relentless discipline. Watch a neat pattern dissolve into structured noise and you stop treating AES like folklore.

The article tours where AES appears, how modes like CBC, CTR and GCM behave, and which operational details — padding, IVs, nonces, authentication, key hygiene — make the difference between “we use AES” as a checklist item and a system that’s actually robust.

About

Italy

A quick note about me.

Portrait of Mario Ciavarella

I'm Mario, a finance graduate who keeps hands-on with code, electronics and AI. I build small but complete things to grow, learn and show what I can do with humility.

What I’m exploring

Ledger visualisations, governance tooling, small creative coding detours.

Current stack

TypeScript, Astro, SQL and Tailwind CSS, with containerized infrastructure on Docker.

  • I define constraints early; it keeps the work grounded and prevents scope drift.
  • I surface unclear parts as soon as I spot them. Ambiguity gets expensive fast.
  • I test with my future self in mind — I'm often the one maintaining the code later.
  • I share progress in small steps; shorter feedback loops make everything smoother.

Contact

Open for new projects and collaborations.

Happy to trade notes or debug a half-finished idea.

I'll be in touch soon.

By submitting this form, you agree that I will use your details to reply to your request. See Privacy Policy.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.