A theoretical reconstruction of the Claude Mythos architecture, built from first principles using the available research literature.

DeepSeek 4 Flash and PRO local inference engine for Metal, CUDA and ROCm

High-performance Transformer inference engine in C++/CUDA — 32k tok/s peak, 87% INT8 Tensor Core ceiling, roofline-validated on RTX 2070

General plug-and-play inference library for Recursive Language Models (RLMs), supporting various sandboxes.

Implement a reasoning LLM in PyTorch from scratch, step by step

Examples of Neuron by me

License Plate Detection using YOLOv8

Eagle: Frontier Vision-Language Models with Data-Centric Strategies

The agent that grows with you

Official repository of the AWS EC2 FPGA Hardware and Software Development Kit

The repo for integrating mini-sglang and AWS Neuron cores

A vector index built on TurboQuant, written in Rust with Python bindings

Diffusion based on single Python file inspired by Andrej Karpathy's microgpt

Agentic malware analysis environment with MCP-connected disassemblers, RE tooling, and structured workflows for Claude Code and Codex CLI.

Enterprise AWS Bedrock proxy gateway with rate limiting and multi-account load distribution, enabling secure access through corporate credentials via standard AWS SDKs like boto3.

Real-time text-to-speech with Qwen3-TTS

Face recognition using Tensorflow

Get started with Recommendation Systems

Runs 405B LLMs on 8GB VRAM

Decky plugin to streamline installation and usage of lsfg-vk; the Lossless Scaling compatibility layer for linux

Building AI slop generator from scratch; No AI tools used, only web articles and papers allowed.

This is a plugin which lets EC2 developers use libfabric as network provider while running NCCL applications.

The best ChatGPT that $100 can buy.

Python SDK for transforming any AI agent into a production-ready application. Framework-agnostic primitives for runtime, memory, authentication, and tools with AWS-managed infrastructure.