Build / Open source

Projects

Projects

Selected Projects

FreeInference Logo

HybridInference

Harvard MadSys Lab
An open-source platform for self-hosted LLM serving. It helps small and medium-sized businesses and research labs deploy their own LLM proxy optimized to minimize cost and latency. The platform powers FreeInference, a free inference service for open-source, research, and education.

01 Platform

hybridInference
Stars

A Hybrid LLM Inference Gateway for Local GPUs and Cloud APIs

hybridInference-cloud-agent
Stars

Cloud Agent Infrastructure in hybridInference

02 Algorithms

RouteWise
Stars

Latency–Cost Optimization for Multi-Provider LLM Routing

03 Data

OpenInfra

A public dashboard for some of the GPUs and serving engine behind FreeInference

OpenData

A dataset hub for LLM serving research

OpenManus-RL Logo

OpenManus-RL

An open-source initiative collaboratively led by Ulab-UIUC and MetaGPT, extending the original @OpenManus project. Inspired by successful RL tuning for reasoning LLM such as DeepSeek-R1, we explore new paradigms for RL-based LLM agent tuning, particularly building upon foundations.
GSoC Logo

ChromiumOS - farfetchd

Selected as a Google Summer of Code 2025 contributor, this project enhances ChromiumOS's farfetchd service by implementing advanced kernel tracing and replaying capabilities, achieving significant reductions in application cold start latency.