
Director, Machine Learning Engineer @ASAPP. AI practitioner, math lover, programmer, tech & security enthusiast. ML, Deep Learning, python, Node & others
New York, USA
Joined August 2009
- Tweets1.7K
- Following1.6K
- Followers2K
- Likes354
NVIDIA is now hosting a bunch of models for free 👀
So I built clauden , run Claude Code on any of them.
e.g:`clauden kimi-k3`
A local LiteLLM proxy bridges the Anthropic ↔ OpenAI APIs. Same agent, same tools, free models (Kimi, DeepSeek, Llama, Qwen)
github.com/gmaggiotti/claude…
ARC-AGI-2 rising from under 5% to 50%+ in months shows rapid progress. Poetiq points to gains from frontier models and smart orchestration without massive compute.
github.com/poetiq-ai/poetiq-…
2
102
Extremely excited about our latest innovation, a state-of-the-art Generative Agent. We are ready to herald the dawn of a new era in machine intelligence and automation. Join us!
shorturl.at/tBKO3
2
165
Excited to see our paper at NeurIPS! 💪
Pretraining LLMs? Scouting 🚄 approximations for nn.linear()?
Coming @ #NeurIPS2023 #WANT WS, @sarin1991 & I explore DYAD - a 2part blocksparse ✖️iable split of weight matrix W.
Pitch tinyurl.com/5687ss7m
Draft tinyurl.com/2rffm7yj
Code tinyurl.com/53rm663u
3
290
Life is short, but the happiness that comes from a perfect slice of pizza lasts forever. 🍕❤️ #PizzaLove #Italy
1
165
As the year comes to a close, I can't help but reflect on all the amazing advancements in the world of tech and AI. From large language models to the release of cutting-edge technologies, it's been an exciting year for us. Here's to an even nerdier & exciting year in 2023. #AI
Gabriel Maggiotti retweeted
How good of a BERT can one get in ONE DAY on ONE GPU?
With all the recent studies about scaling compute up, this paper takes a refreshing turn and does a deep dive into scaling down compute.
It's well written, stock full of insights. Here is my summary and my opinions.
🧶 1/N
I just got a chance to work with PaLM and I'm amazed by its capabilities!
Someone implemented a framework to allow you to train PaLM (Google’s 540B parameter LM) using the same reinforcement learning strategy as ChatGPT!
Open source strikes again.
To start, here’s the part where you initially pre-train the model:
2
352
We are growing our ML Engineering team!!! If you'd like to work on cutting edge machine learning and natural language processing (NLP) challenges and help us to turn our research into meaningful AI-Native products, please DM me.…lnkd.in/dz7ZKBuT lnkd.in/dW-kiS24
Gabriel Maggiotti retweeted
It’s the death of canned responses. We launched AutoCompose to provide #CustomerService Agents with personalized guidance and tailored language that automates 70% of digital conversations to shorten the time it takes customers to get Qs resolved.
youtu.be/KEdNWAyM5R0



















