This image explains to me very well the importance of Data-to-text Natural Language Generation.

I'm excited to share that I will be extending my internship to continue working on a really inspiring project at Amazon Science 🎉 Very happy to work with Anurag Beniwal, Arjun Seshadri, Tian Wang, and the rest of the team!

Happy International Programmers' Day! This is celebrated on the 256th day of the year, which is today. ❤️ all of you writing or learning to write software for other people to use!

It is funny because is true..

I'm a proud advisor to 4 of these students! not only diverse but also very smart and talented Xinnuo Xu Karin Sevegnani Malvina Nikandrou Nikolas V. 🦋

SIGDIAL 2022 SIGdial is at Heriot-Watt University Edinburgh, Scotland. Organised by the Interaction Lab and National Robotarium @NRobotarium : 7-9 September 2022 - save the date! #SIGdial2022 2022.sigdial.org

Lessons learned debugging ML models:

😅Chinchilla drawing analogies between concepts:

Did you know that a UK PhD stipend is so little that it puts recipients below the absolute poverty line? UK PhD stipend (outside London): £16,062 Absolute poverty line (60% of 2011 median income): £16,883 No one does research for the money but surely we can do better than this?

Excited to share that the paper I've worked on at Amazon Science during my internship: "Contrastive Learning for Interactive Recommendation in Fashion" has been accepted at SIGIReCom! 🎉 Work in collaboration with Anurag Beniwal and the rest of the team. ✨

Just how much have language models grown in the last 4 years? Let's have a look. In 2018, the puny BERT “large” model premiered with a measly 354M parameters. It can be trained on a single 8xA100 node in 5 days. That costs $2K on AWS - almost free by LLM standards! 🧵

I will be presenting this work Friday the 15th at ~2pm (CET) ✨ Camera-ready version available here: sigir-ecom.github.io/ecom22Papers/p…

📢 3 new research-focused faculty positions Heriot-Watt University computer science. We're looking to strengthen our team in #NLProc and #ConversationalAI. DM me if interested

Finishing day 2 with the #SIGdial banquet, and it is wonderful to see everyone! We hope you all enjoy the meal before... the ceilidh! The band are getting ready 🎉 #SIGdial2022 Ghillie Dhu #Edinburgh

with huge thanks for work of Angus Addlesee Christian Dondrup Daniel Hernandez Neeraj Cherakara Nancie Gunson Weronika Sieinska, Marta Romeo, Sabrina Chiesurin Karin Sevegnani vevake Balaraman Yannis Konstas Yiannis Papaioannou Arash Eshghi Alessandro Suglia @joselpart Katie Baker Claudio Greco Marco Sobrevilla

Special thanks to Giuseppe Fiameni Giuseppe Fiameni and the team at NVIDIA for visiting University of Bristol to give us a MaVi Seminar @BristolUniEng on Scalable AI - Scaling Training for DNNs. Looking forward to further collaborations when Isambard-AI is live!

I’m excited to share that I have joined NVIDIA AI as a Senior Solutions Architect! ✨

🚨 New top scorer on the BALROG LLM leaderboard! DeepSeek R1, evaluated via NVIDIA’s NIM API, takes the lead. Its strong reasoning capabilities help it make solid progress on BALROG’s toughest tasks — the best performance we’ve seen so far! 🧠⚔️

We are transforming #AI model and benchmark evaluation with NVIDIA NIM by providing seamless integration with OpenAI-compatible APIs and optimized GPU-powered inference engines. 💡 Key benefits include: ✅ Effortless deployment of large-scale LLMs like DeepSeek-R1. ✅

🚀 Excited to share two new blog posts! 1️⃣ Profiling LLM training on NVIDIA Grace Hopper: developer.nvidia.com/blog/profiling… 2️⃣ Advanced optimization strategies: developer.nvidia.com/blog/advanced-… Unlock better performance for your LLM workflows! #AI #NVIDIA #LLM #GraceHopper