I’m aPrincipal ML Engineer atLevel AI, where I focus on building and scaling large language models (LLMs) specifically for conversational AI. With over four years of experience in applied AI and research, I’ve worked extensively on end-to-end solutions in NLP and ML systems.

Before Level AI, I worked as aData Scientist atMastercard AI Garage, where I developed AI models to enhance transaction security and intelligence. I graduated in 2020 with a degree in Computer Science from theIndian Institute of Technology (BHU) Varanasi.

My technical interests include Natural Language Processing, ML Systems Engineering—including CUDA and Triton for high-performance computing, Privacy-preserving ML, and Cryptography.

I’m always working on side projects, many of which involve implementing and experimenting with ideas from research papers, efficient kernels and other low-level stuff in LLM training/inference regime. You can find these projects here.

• 𝕏 Twitter/X:@shreyansh_26
• 👥 LinkedIn:Shreyansh Singh
• 💻 Website:https://shreyansh26.github.io

• Deep dive into CUDA Scan Kernels: Hierarchical and Single-Pass Variants
• Paper Summary #14 - Physics of Language Models: Part 3.1, Knowledge Storage and Extraction
• Understanding Multi-Head Latent Attention (MLA)
• Deriving the Gradient for the Backward Pass of Layer Normalization
• Notes from GTC’25: CUDA Techniques to Maximize Compute and Instruction Throughput

PinnedLoading

1. Annotated-ML-PapersAnnotated-ML-PapersPublic

   Annotations of the interesting ML papers I read

   275 27

2. FlashAttention-PyTorchFlashAttention-PyTorchPublic

   Implementation of FlashAttention in PyTorch

   Python 181 23

3. Extracting-Training-Data-from-Large-Langauge-ModelsExtracting-Training-Data-from-Large-Langauge-ModelsPublic

   A re-implementation of the "Extracting Training Data from Large Language Models" paper by Carlini et al., 2020

   Python 39 7

4. Speculative-SamplingSpeculative-SamplingPublic

   Implementation of Speculative Sampling as described in "Accelerating Large Language Model Decoding with Speculative Sampling" by Deepmind

   Python 108 14

5. LLM-SamplingLLM-SamplingPublic

   A collection of various LLM sampling methods implemented in pure Pytorch

   Python 27 3

6. Linux-Malware-Detection-ResearchLinux-Malware-Detection-ResearchPublic

   A collection of Linux Malware Detection projects (research paper implementations) done by me.

   Jupyter Notebook 11 2