Energy-efficient and SLO-aware disaggregated LLM serving system.

A multi-trigger trojan attack framework for LLM-assisted robot task planners.

Method for compressing large VLMs into lightweight students using multi-teacher ensembles.

A system for certifying that agentic AI actions satisfy formal policies using cryptographic proofs.

Open-source AI framework for building production-ready LLM applications and RAG pipelines.

An open data curation pipeline and dataset for training broadly capable agentic LLMs.

Decentralized training protocol partitioning models into independent blocks for distributed optimization.

Resource-efficient KV-cache compression framework for GQA-based LLMs.

A self-hosted realtime trip planner with LLM integration and collaboration.

Guide and code for an ESP8266-based power management system using MQTT for homelab servers.

Unified real-time end-to-end vision models for object detection, segmentation, and classification.

Auto-updated IP ranges for cloud providers, CDNs and bots with config generators.

A Git replacement designed for AI agents to handle version control more efficiently.

A based-on-Alpine Linux distribution for old smartphones and tablets.

An open source memory server for AI tools utilizing Cloudflare's free tier for semantic recall.

A tool to use any Linux distribution inside your terminal with full integration

An ATProto Personal Data Server (PDS) that runs on Cloudflare Workers.

An open-source game engine for 2D and 4D games.

A Bi-Anchor Interpolation Solver for accelerating generative modeling with minimal training cost.

A brain-inspired always-on agent memory system for proactive assistants.