Skip to content
View codiceSpaghetti's full-sized avatar
๐ŸŽฏ
Focusing
๐ŸŽฏ
Focusing

Block or report codiceSpaghetti

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.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
codiceSpaghetti/README.md

๐Ÿ‘‹ Hi there, I'm Alessio ๐Ÿ‡ฎ๐Ÿ‡น

๐Ÿ‘จโ€๐Ÿ’ป About me in O(1)

I'm an AI Researcher & Engineer focused on multilingual large language models. ๐Ÿค–
At Translated, I helped build from scratch Lara, a machine translation LLM that powers content in real-time for Airbnb, Uber, Shopify, and others, reaching over 200 million users. I've scaled training to 1,000+ GPUs and expanded MT coverage to 201 languages. My work has consistently aimed at one thing: making LLMs speak your language. ๐Ÿ”Š๐ŸŒ

Working nights and weekends, I co-founded an AI startup building an image geolocalization system as part of a three-person founding team. That experience taught me how everything can be uncertain, fast, and change in a blink. We were figuring things out as we went, but growing quickly. It also brought a few bitter lessons about people...

I'm a voracious learner. I constantly push to improve how I think, train, and build. I can't imagine a life without some form of training, whether it's GPUs ๐Ÿ–ฅ๏ธ, gym plates ๐Ÿ‹๏ธ, or marathon prep ๐Ÿƒ.

Still iterating.

๐Ÿ”ง Some tools I Use

Category Tools
Languages & Core Python Java Bash PyTorch Transformers Hugging Face
Data & Processing FastText Pandas NumPy SQL MongoDB FAISS Tiktoken spaCy
Training & Inference FSDP DeepSpeed PEFT Axolotl vLLM FlashAttention TensorRT ONNX
Monitoring & Deploy Slurm Singularity Docker MLflow TensorBoard W&B

๐Ÿ“ž Reach Out

LinkedIn ย  X ย  Gmail

Pinned Loading

  1. T4SA-2.0 T4SA-2.0 Public

    This project creates the T4SA 2.0 dataset, i.e. a big set of data to train visual models for Sentiment Analysis in the Twitter domain using a cross-modal student-teacher approach.

    Jupyter Notebook 3 2

  2. FakeNewsDetection FakeNewsDetection Public

    This project explores both Transfer Learning and Feature Extraction for obtaining contextual word embeddings using BERT-family model to solve a problem related to the Fake News Detection task, i.e.โ€ฆ

    Jupyter Notebook 1

  3. numpyGPT numpyGPT Public

    A from-scratch GPT built with NumPy and Pythonโ€™s standard library. No autograd, no frameworks: every layer is re-implemented with its own forward and backward pass. Gradients are computed manually,โ€ฆ

    Python

  4. UniMusic UniMusic Public

    UniMusic is an application that aims to simplify and enhance the song discovery and management experience for its users. With this platoform users can find and access official sources of songs on tโ€ฆ

    Java 1

  5. VoiceVibes VoiceVibes Public

    VoiceVibes presents a solution for speech emotion recognition, featuring six innovative AI model architectures designed to accurately categorize emotional expressions conveyed through speech.

    Jupyter Notebook 3 1

  6. FederatedDBSCAN FederatedDBSCAN Public

    Implemention of the DBScan clustering algorithm with Federated Learning setting. The goal is to enable distributed clustering of data across multiple nodes in a decentralized network, preserving thโ€ฆ

    Python 6 1