Swayam Singh
"Research at Microsoft Research, Contributing to Low-Level Numpy Enhancements"
Hello Internet wanderer! I’m Swayam Singh, an inadvertent wizard in the ever-expanding realm of data science and machine learning. And no, I haven’t been to Hogwarts, but if they ever offer a course in Deep Learning, I’d be the first in line – think of it as the Muggle’s Defense Against the Dark Arts!
On the professional front, I’ve proudly worn the hats of Data Science Intern, Machine Learning Engineer, Open Source Research Engineer, and now, a Research Fellow at Microsoft Research with the AI4Code team. Additionally, I contribute to the open-source community through my work with Quansight Labs on Numpy. My ventures in the tech realm encompass pivotal roles from delving into customer retention and mastering the art of demand forecasting, to streamlining machine learning pipelines and orchestrating robust distributed systems. My hands have danced on custom CUDA kernels, ensuring that MLOps is not just a buzzword, but a seamlessly integrated practice, and that distributed processing is executed with finesse and efficiency. Furthermore, my research escapades allowed me to rub shoulders with some of the brightest minds, resulting in state-of-the-art Code-gen models—twice, to be precise.
Hobbies & Fun Bits:
- 📺 Binge-watch Anime & Groove on the soulful vibes of POP+R&B🥂.
- ⚽ Show off some fancy footwork on the football pitch, channeling inner Cedric Diggory vibes.
- 📚 Occasionally geek out over Quantum physics—yes, it’s nerdy, but hey, you’re in my ‘Interests’ section.
My journey in tech has been as thrilling as a Quidditch match, with dizzying highs and the occasional Bludger-like setback. In a twist straight out of a manga, I found myself amidst the prodigious minds of Amazon ML Summer School. Talk about a steep learning curve!
I’ve embarked on various projects, some of which seemed straight out of a Shonen plotline. If you’re seeking someone who’s ever-curious, can appreciate a good ‘Expelliarmus’ joke, and can pivot from discussing convolutional networks to the latest anime plot twist, you’ve found your guy. Let’s embark on this enthralling quest together, and who knows? We might just stumble upon the Philosopher’s Stone of technology—or at least code that doesn’t require a Time-Turner to debug!
News
Jul 4, 2024 | 🚀 New Paper and model out Narrow Transformer: Starcoder-Based Java-LM For Desktop |
---|---|
Jul 1, 2024 | 🚀 Joined Quansight Labs as OS Engineer (Numpy), working to implement and deploy an extended Floating Point Quad-Precision Data Type |
Jul 1, 2024 | 🚀 Joined Microsoft Research as Research Fello, working with AI4Code team |
Jun 1, 2024 | 🚀 Became Kaggle Competition Expert: Check Out the Profile. |
Mar 16, 2024 | 🚀 Gave a talk on “Mamba Zero To Hero” at Cohere for AI. Discussing the past, present and future of State Space Models |
Feb 19, 2024 | 🚀 Launched a new project MIRA - Multimodal Image Reconstruction with Attention is a transformer (Encoder-Decoder) based architecture for Text / Image to 3D reconstruction just using single 2D image of object within seconds. Check it out 🥂 |
Latest Posts
Oct 10, 2024 | Understanding Perplexity |
---|---|
Sep 30, 2024 | Numpy QuadDType: Quadruple Precision for Everyone |
Oct 14, 2023 | Self-Attention Mimicking Gradient Descent |
Publications
- OctoPack: Instruction Tuning Code Large Language Models2023
- StarCoder: may the source be with you!2023
- Narrow Transformer: Starcoder-Based Java-LM For Desktop2024