Post
125
🧠 Just shipped ML Research Assistant on HuggingFace Spaces!
An AI research assistant powered by Qwen2.5-7B with 17 auto-routing tools. Ask in plain English — it figures out the right tool automatically.
🔧 What it can do:
• Compare LLMs on real benchmarks (MMLU, GSM8K, HumanEval, HellaSwag)
• Search ArXiv papers and find GitHub implementations
• Explore model benchmarks with real F1 and Accuracy scores
• Live AI news feed — click any headline to ask the assistant about it
• Web search, code generation, HuggingFace model and dataset search, and more
💡 Try these:
→ compare gpt-4o vs claude 3.5 sonnet on coding
→ show me the overall llm leaderboard
→ find papers with code for transformer attention
→ trending ml repos on github
⚙️ Built without LangChain — I wrote the router from scratch so the model only gets called once per query. That matters on ZeroGPU where GPU time is limited.
⚠️ Still a v1 — handles one query at a time, cold start delay on first load. Multi-query support and streaming are coming next.
🔗 Abdourakib/ml-research-assistant
📁 https://github.com/Abdrakib/ml-research-assistant
Would love feedback from the community — what would you add?
An AI research assistant powered by Qwen2.5-7B with 17 auto-routing tools. Ask in plain English — it figures out the right tool automatically.
🔧 What it can do:
• Compare LLMs on real benchmarks (MMLU, GSM8K, HumanEval, HellaSwag)
• Search ArXiv papers and find GitHub implementations
• Explore model benchmarks with real F1 and Accuracy scores
• Live AI news feed — click any headline to ask the assistant about it
• Web search, code generation, HuggingFace model and dataset search, and more
💡 Try these:
→ compare gpt-4o vs claude 3.5 sonnet on coding
→ show me the overall llm leaderboard
→ find papers with code for transformer attention
→ trending ml repos on github
⚙️ Built without LangChain — I wrote the router from scratch so the model only gets called once per query. That matters on ZeroGPU where GPU time is limited.
⚠️ Still a v1 — handles one query at a time, cold start delay on first load. Multi-query support and streaming are coming next.
🔗 Abdourakib/ml-research-assistant
📁 https://github.com/Abdrakib/ml-research-assistant
Would love feedback from the community — what would you add?