Open source implementation and extension of Google Research’s PaperBanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation.

SQL-like query language and CLI for Qdrant vector search engine

This repo is meant to serve as a guide for Machine Learning/AI technical interviews.

DFlash & TurboQuant in llama.cpp with up to 3x faster generation and 7.5x more KV cache in same VRAM

Advanced prompt injection defense system for AI agents. Multi-language detection, severity scoring, and security auditing.

HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal

A fast type checker and language server for Python

Fastino's LLM guardrail

A curated list of resources for learning and exploring Triton, OpenAI's programming language for writing efficient GPU code.

Minimal keyword extraction with BERT

ML algorithms implemented and derived from first-principles in Jupyter Notebooks and NumPy

Modal-style sandbox API on top of Hugging Face Jobs

DFlash: Block Diffusion for Flash Speculative Decoding

Codebase for LLM Textual Hallucination Benchmark

Development repository for the Triton language and compiler

TokenSpeed is a speed-of-light LLM inference engine.

Pure Rust Inference Engine

MedSafetyBench: Evaluating and Improving the Medical Safety of LLMs, NeurIPS 2024

hardware implementation of transformers running microgpt at 50k+ tkps

Domain-specific GLiNER fine-tuning pipeline for sports NER. GLiNER is a bidirectional transformer encoder (DeBERTa-v3 based) for zero-shot named entity recognition — I'm fine-tuning it on a custom …

OpenAI Privacy Filter

A framework for few-shot evaluation of language models.

DuoGuard: A Two-Player RL-Driven Framework for Multilingual LLM Guardrails

Restore heading hierarchy in markdown documents using a fine-tuned 0.6B parameter LLM.

Claude skill for finding ML research papers.

KV cache compression via block-diagonal rotation. Beats TurboQuant: better PPL (6.91 vs 7.07), 28% faster decode, 5.3x faster prefill, 44x fewer params. Drop-in llama.cpp integration.