Skip to content
View cool-develope's full-sized avatar
🏠
Working from home
🏠
Working from home

Block or report cool-develope

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
cool-develope/README.md

Hi, I'm an AI Engineer 👋

I build production-oriented AI systems and autonomous agents — from retrieval and deep research to coding agents, tool use, evaluation, and secure execution.

My focus is turning LLM capabilities into reliable software systems.

🤖 What I work on

  • AI Agents — planning, tool use, multi-agent workflows, memory, and orchestration

  • LLM Systems — model routing, structured generation, inference, and evaluation

  • RAG & Deep Research — heterogeneous ingestion, retrieval, reranking, citations, and long-context workflows

  • Coding Agents — repository understanding, SWE-bench workflows, code generation, and execution

  • Computer-Use Agents — browser/computer interaction and multimodal agent evaluation

  • Agent Infrastructure — sandboxed execution, observability, pipelines, storage, and deployment

🚀 Selected Projects

🔬 Material Research

Self-hosted AI research system for turning heterogeneous research materials — papers, source-code projects, datasets, and archives — into searchable and citable knowledge.

RAG · Deep Research · Document Intelligence · Python · Dagster · PostgreSQL · MinIO · Qdrant

⚔️ Duel Agents

Multi-model agent routing and evaluation system focused on selecting effective model responses while controlling inference cost.

LLM Routing · Agents · Evaluation · Model Selection

💻 SWE-bench Agent Experiments

Experiments in autonomous software engineering and repository-level problem solving using SWE-bench.

Coding Agents · SWE-bench · Code Generation · Evaluation

🖥️ Computer-Use Agent Evaluation

Exploration of multimodal computer-using agents and environments for evaluating tool-driven interaction.

Computer Use · Multimodal Agents · Agent Evaluation · MCP

🔐 Agent Sandbox

Secure execution infrastructure for AI agents using WebAssembly-based isolation.

Agent Infrastructure · Sandboxing · WASM · Secure Execution

🧠 Engineering Interests


LLM Agents        ████████████████████

RAG / Retrieval   ████████████████████

AI Engineering    ████████████████████

Agent Evaluation  ██████████████████

LLM Infrastructure██████████████████

MLOps             ████████████████

🛠️ Stack

AI / LLM

LLMs · RAG · Agents · Embeddings · Reranking · MCP · SWE-bench

Backend

Python · FastAPI · Pydantic · AsyncIO

Data & Retrieval

PostgreSQL · Qdrant · MinIO · Redis

AI Infrastructure

Dagster · Langfuse · Ollama · Docker

🎯 Current Focus

Building AI systems that can:

understand → retrieve → reason → use tools → execute → evaluate → improve

with an emphasis on reliability, observability, evaluation, and production deployment.

Pinned Loading

  1. material-research material-research Public

    Turn research zips into searchable, citable materials. Split papers, code, and datasets; index them; ask Deep Research questions that cite pages and source lines. Self-hosted (Postgres, MinIO, Qdra…

    Python

  2. duel-agents duel-agents Public

    Use, extend, and ship with Duel Agents: the IDE-native routing layer that runs prompts against multiple models and picks the cheapest answer that still wins.

    TypeScript

  3. hack-swebench hack-swebench Public

    Achieved a 52.4% resolve rate on SWE-bench Lite

    Python

  4. mcp-world mcp-world Public

    MCPWorld: A Multi-Modal Test Platform for Computer-Using Agents (CUA)

    Python

  5. wasm-sandbox wasm-sandbox Public

    A monorepo containing WASM sandbox packages for secure code execution in AI agent environments.

    TypeScript

  6. agent-stacklit agent-stacklit Public

    One command makes any repo AI-agent-ready. No server, no setup.

    Go