Projects

Overview: This repository contains optimized JAX code for OpenAI's Whisper model, achieving up to 70x speed-up on TPU compared to PyTorch code. It is compatible with CPU, GPU, and TPU, and can be run both standalone and as an inference endpoint. Key features: - Optimized JAX code for Whisper model - Compatible with CPU, GPU, and TPU - Standalone and inference endpoint options - Half-precision computation for speedup - Batched audio processing for significant speed-up Tech stack: JAX Why it matters: This implementation provides a significant performance boost, making it the fastest Whisper model available, especially on TPU devices. Repository: sanchit-gandhi/whisper-jax GitHub stars: 4677

### Overview & Goals **ContextForge** is an open-source registry and proxy designed to federate tools, agents, and APIs into a unified endpoint for AI clients. It provides centralized governance, discovery, and observability across your AI infrastructure. Key features include: - **Tools Gateway**: MCP, REST, gRPC-to-MCP translation, and TOON compression. - **Agent Gateway**: A2A protocol, OpenAI-compatible and Anthropic agent routing. - **API Gateway**: Rate limiting, authentication, retries, and reverse proxy for REST services. - **Plugin Extensibility**: 40+ plugins for additional transports, protocols, and integrations. - **Observability**: OpenTelemetry tracing with Phoenix, Jaeger, Zipkin, and other OTLP backends. **ContextForge** runs as a fully compliant MCP server, deployable via PyPI or Docker, and scales to multi-cluster environments on Kubernetes with Redis-backed federation and caching. ### Key Features - **Tools Gateway**: MCP, REST, gRPC-to-MCP translation, and TOON compression. - **Agent Gateway**: A2A protocol, OpenAI-compatible and Anthropic agent routing. - **API Gateway**: Rate limiting, authentication, retries, and reverse proxy for REST services. - **Plugin Extensibility**: 40+ plugins for additional transports, protocols, and integrations. - **Observability**: OpenTelemetry tracing with Phoenix, Jaeger, Zipkin, and other OTLP backends. ### Tech Stack - Python - Redis ### Why It Matters -

Overview: Align Anything is a comprehensive framework designed to align any modality large models with human intentions and values, enabling the fine-tuning of various multi-modal models and the implementation of different alignment methods. Key features: - Highly modular framework for easy customization. - Various modality model fine-tuning capabilities. - Different alignment methods including SFT, DPO, PPO, and others. - Multi-modal CLI for image, audio, and video modalities. - O1-like training based on DollyTails. - Rule-based RL encouraged by Deepseek-R1. Tech stack: Python. Why it matters: Align Anything provides a powerful tool for researchers and developers to fine-tune large language models across multiple modalities, making it easier to achieve better alignment with human values and intentions.