Collaborate

I occasionally collaborate with researchers, founders, and technical teams on problems close to the boundary between AI research and engineering.

I am most useful when the problem is technically dense, ambiguous, and not yet well served by a generic implementation recipe: an agentic system that needs a clear harness, a research prototype that needs to become reproducible, an evaluation that needs sharper failure modes, or a scientific AI idea that needs to move from paper logic to working infrastructure.

Areas

• Agentic AI systems, harness design, tool use, memory, and workflow infrastructure
• LLM evaluation, failure analysis, and benchmark design
• Research prototype design and implementation
• Paper-to-system translation for AI research ideas
• Technical review of AI product or research directions
• Scientific intelligence and foundation models for science

Formats

• Short technical advisory conversations
• Architecture or research-plan review
• Prototype, benchmark, or evaluation sprint
• Research engineering collaboration
• Longer-term collaboration when there is strong fit

Good Fit

This is likely a good fit if you need someone to reason through a technical direction, identify failure modes, design an experiment, or turn an AI research idea into a working system.

I am usually not the right person for generic chatbot integration, marketing automation, SEO content generation, or projects where the main requirement is fast outsourced implementation.

Contact

If there is a possible fit, email me at di.zhang@ustc.edu with a short note about the problem, timeline, and what kind of outcome would be useful.