Skip to content
View arash7m's full-sized avatar
🏹
🏹

Block or report arash7m

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. 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
arash7m/README.md

Hi, I'm Arash. AI and LLM engineer, developer and designer.

AI / LLM engineer with an M.Sc. in Computer Engineering (Artificial Intelligence) from the University of Padova.
I build LLM pipelines that are grounded, validated and measured, and small apps that respect the people using them.
Background in networking, cybersecurity and web development. Long-time fan of computer music and digital audio.

LinkedIn Open to AI / LLM engineering roles Based in Padova, Italy

Featured projects

M.Sc. thesis · 2025

An end-to-end LLM pipeline that analyses and documents legacy music code: 64 Csound projects, 36,000+ lines.

  • 83% F1 on factual code-comprehension questions
  • Full-codebase context beat chunked baselines by 14 F1 points on signal-flow mapping
  • JSON Schema–validated output, grounded in the audio through librosa spectrograms
  • Multi-judge evaluation harness (GPT-4, Claude, Gemini) that measures where judges disagree
  • Reproducible Docker image; corpus, pipeline and evaluation released as open source

Python LLaMA 3.3 via Groq librosa Docker

Open source · 2026

Small, private tools for trans and gender-diverse people, as an installable web app that works offline.

  • Real-time pitch detection (YIN) and resonance estimation (LPC formants), in the browser
  • Voice games, a guided daily warm-up and live graphs
  • Dose tracking, lab charts, a journal, body measurements and a doctor summary
  • No server and no tracking: everything stays on the device, with optional AES-GCM encryption

JavaScript Web Audio PWA AGPL-3.0

What I work with

AI / ML: large language models · prompt engineering · LLM evaluation (LLM-as-judge) · structured output with JSON Schema · deep learning · reinforcement learning

PyTorch TensorFlow Hugging Face Transformers scikit-learn librosa Unity ML-Agents

LLM APIs:

Groq (LLaMA 3.3) OpenAI GPT-4 Anthropic Claude Google Gemini

Languages and tools:

Python C# JavaScript TypeScript HTML and CSS Node.js Docker Git Unity

Networking and security: routing and switching · firewalls · security monitoring · vulnerability assessment

Background

  • 🎓 M.Sc. Computer Engineering, Artificial Intelligence, University of Padova (2022–2025). Thesis: LLMs for Understanding and Preserving Historical Musical Codes.
  • 🎓 B.Sc. Software Engineering (2016–2021). Thesis: reinforcement-learning agents trained with Unity ML-Agents in a custom game environment (C#, Python).
  • 🌐 Network engineer (2021–2022): network infrastructure plus front-end, back-end and full-stack web projects in JavaScript, TypeScript and Node.js, in an agile team.
  • 🛡️ Cybersecurity intern (2021): security monitoring and vulnerability assessment with Python tooling.

🗣️ English · Persian · Italian (learning)  ·  🎧 computer music & digital audio · 🎮 game design · 🎨 web design · ✍️ creative writing

Pinned Loading

  1. Tendril Tendril Public

    Small, private tools for trans and gender-diverse people. Everything stays on your phone

    JavaScript 2

  2. CSoundAnalyzer CSoundAnalyzer Public

    Analyzing CSound Projects using LLMs

    Csound 1

  3. 911_Calls 911_Calls Public

    Data Capstone Project: the 911 calls of Montgomery County, PA

    Jupyter Notebook 1

  4. AdClicker--LogisticReg AdClicker--LogisticReg Public

    Using fake data to evaluate whether a user clicked on an ad or not

    Jupyter Notebook 1

  5. Car_Game_platformer_3D Car_Game_platformer_3D Public

  6. CV-UNIPD CV-UNIPD Public

    Cv submissions 23-24

    C++