Pashmina Cameron

👉 Apply now or DM to learn more. Principal Researcher - Redmond, US: lnkd.in/eCqaniDx Senior Applied Scientist - Redmond, US: lnkd.in/e9KXazP2 Research Intern - Redmond, US: lnkd.in/e_CSAa4d

🚀 We’re hiring #researchers, #appliedscientists and #interns passionate about #Agents and Multimodal #LLMs/#SLMs. Shape the next wave of transformative experiences. If you love building systems that think, reason, and act-your next agent mission awaits! #Microsoft

A slice 🍰 of what we do. It's a multi-tier cake🎂 with a cherry 🍒 on top. blogs.windows.com/windowsexperie…

We have #machinelearning #intern positions available at #microsoft Applied science group in #LLMs, #diffusion models, #multimodal models and #efficient AI. Apply at: jobs.careers.microsoft.com/global/en/job/…. Learn more about our work here: youtu.be/ET_iZnthP7o?fe…

Sharing a few highlights from the week. Grateful for the teamwork that made it all possible. Each step forward fuels our passion for the journey ahead!

My team helped create a new device category. Under the hood of the Copilot+ PC's are a screaming NPU chip and new #AI models in a slim, portable package with amazing battery life. I explain a bit more in my #MSBuild talk. youtube.com/watch?v=Ov23WF… #AIPC #Copilot+PC #Surface

🤩We have an exciting opportunity to join TVG as a PDRA in Machine Learning, this position is in collaboration with Microsoft! Apply here: rb.gy/w7aum3 Deadline: 4th April @oxengsci @microsoft @Windows @surface Applied Sciences Group

Microsoft Applied Sciences Group is hiring a Principal Scientist to work with LLMs and multimodal models with an eye towards on-device deployment. Come help us bring the goodness of LLMs and Multimodal models to everyone. jobs.careers.microsoft.com/global/en/job/…

We're excited to announce the launch of Phi-3, a groundbreaking family of small language models that outperform larger models on a range of benchmarks. Learn how these small language models trained on high-quality data are doing more with less: msft.it/6010YHP32

Game on! huggingface.co/microsoft/Phi-…

phi-3 is here, and it's ... good :-). I made a quick short demo to give you a feel of what phi-3-mini (3.8B) can do. Stay tuned for the open weights release and more announcements tomorrow morning! (And ofc this wouldn't be complete without the usual table of benchmarks!)

Exciting #Microsoft Applied Sciences Group with Phil Torr @OxfordTVG collaboration with funded #postdoc position leveraging efficient #LLMs and #RL for holistic video understanding.

@hmemcpy He's making a list He's checking it twice Santa should've fused the two traversals into one

Phi-2 numbers, finally! We're seeing a consistent ranking: phi-2 outperforms Mistral 7B & Gemini Nano 2* (*on their reported benchmarks) and is roughly comparable to Llama 2-70B (sometimes better, sometimes worse). Beyond benchmarks, playing with the models tells a similar story.

Today, we share our teams’ latest contributions, Phi-2 and promptbase. Phi-2 outperforms other existing small language models, yet it’s small enough to run on a laptop or mobile device. msft.it/6040ipYH6

Really looking forward to presenting our paper "HuManiFlow: Ancestor-Conditioned Normalising Flows on SO(3) Manifolds for Human Pose and Shape Distribution Estimation" at CVPR 2023! w/ @IgnasBud and @robertocipolla Paper + code: github.com/akashsengupta1… 1/3

In "Are VAEs Bad at Reconstructing Molecular Graphs?" we see how common auto-encoders of drug-like molecules perform at recovering their inputs. Answer: not too well, but we actually obtain the best property optimization results when reconstruction is low. arxiv.org/abs/2305.03041

Another year & another VizWiz workshop at @CVPR!🤩This year we've got 4 exciting #AI challenges to better assistive technologies for the blind/low-vision community: VQA, VQA Grounding, Few-Shot Object Recognition & Salient Object Detection. bit.ly/3GGM34l. Deadline 5 May

@DataChaz This assumes that AIs will be freely available to all. Will they be? Who will pay for this? Who ensures equitable access to the technology? I’m all in on AI for a better world as long as it’s a better world for everyone and not just the privileged.

The SUPR model is our latest generative 3D human. SUPR is learned from 1.2M 3D scans, more than any other 3D human. This makes it highly accurate and realistic. Unlike previous models, SUPR includes a realistic foot that deforms with contact. (1/8)