uber/adr
ADR secures enterprise AI agents through observability, security benchmarking, and threat detection. Deployed at Uber.
01 / MACHINE INTELLIGENCE
Models you can run. Agents you can inspect. Tools you can build on.
← All topicsADR secures enterprise AI agents through observability, security benchmarking, and threat detection. Deployed at Uber.
Open the project to explore the idea and the code.
ReflectVLN: Training Vision-Language Navigation Agents with Self-Reflective Reasoning
Your coding agent starts every session blank. Rekal is the memory your team is missing
Open the project to explore the idea and the code.
SIRIN is a Python library for detecting and analyzing contextual inconsistencies in systems that use LLMs
ARIS ⚔️ (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in — works with Claude Code, Codex, OpenClaw, or any LLM agent.
Give your agent a computer 👾.
🚀 The definitive roadmap to becoming a Forward Deployment Engineer (FDE). Master AI Agents, Enterprise Data Architecture, and Strategic Consulting. Bridging the gap between HQ and the field. Inspired by the "Delta" role at Palantir, OpenAI, and Scale AI.
The most powerful local music generation model that outperforms almost all commercial alternatives, supporting Mac, AMD, Intel, and CUDA devices.
Qwen-CUA: Native Computer Use for (Almost) Everything — a screenshot-driven agent that operates computers with keyboard and mouse, jointly developed by the Qwen Team and XLang Lab.
Turn Codex threads into an organization of long-lived domain agents.
A 10-week, 30-minutes-a-day roadmap for LLM inference serving and optimization. vLLM, SGLang, quantization, speculative decoding, benchmarking.
macOS menu bar app for ds4: fast DeepSeek V4 Pro/Flash, with 1M context.
An AI-native office suite for macOS and Windows: word processor, spreadsheet, presentations, and PDF.
A minimal autonomous code-agent CLI built with Qwen Code
Find out which subscriptions are one prompt away from free. caniuse.com, but for replacing SaaS with one AI prompt.
Counterfactual replay and causal attribution over LLM-agent trajectories: find which step actually caused a failure by intervening on it.
Team of marketing agents built on eve.
Composable Pi coding agent with MCP, LSP, agent chains, prompt presets, and local eval telemetry
Claude Code Marketplace.
Local AI powered red teamer on a phone.
Docker images for LLM inference (SGLang + vLLM) on NVIDIA Blackwell GPUs (SM120, CUDA 13.2)
Tools for Agents.