Xuhui Zhan · Applied Scientist
Curiosity,
built into
systems.
I turn research questions into working AI. My work connects multimodal learning, recommendation systems, and the infrastructure that brings them to life.
A connected practice
Follow a point. Explore the work.
01 / 06Treverse RecsysCurrently building at Treverse LLC ↗
Research ↔ production Nashville, TN
01 / Selected work
Ideas made real.
A selection of research and engineering projects, from the first question to the decisions behind the system.
Treverse Recsys
Personalized recommendations, from temporal graph learning to real-time serving and controlled experiments.
Read case study
Inverse-LLaVA
Rethinking how vision and language meet, with no alignment pre-training and 45% fewer training samples.
Read case study
AI for Negotiation
Turning negotiation transcripts into research data, reducing annotation cost from $5,000 to $3 per transcript.
Read case study
Reading history through materials
Computer vision meets archaeology: classifying ancient mortar particles with 97% accuracy.
Read case study02 / How I work
Research and production,
in one loop.
A useful model is only part of the story. I follow the question through evaluation, deployment, and what the system teaches us next.
Method / 05
Each stage opens the project that best demonstrates it.
Frame the question.
Ancient Mortars structures roughly 10 million particle images for comparison.
Fit the representation.
Inverse-LLaVA maps text into visual space inside the language model.
Make the comparison legible.
Negotiation transcript coding reports a cost drop from more than $5,000 to $3.
Publish a reviewable release.
FORGE defines six contracts across distinct ML runtimes.
Return field evidence.
Treverse joins controlled exposure with reproducible learning and serving.
03 / Experience
A path through
research & practice.
Beyond the models
A curious mind.
A human behind the work.
My interests take me from language models to archaeological materials. Away from the work, there’s usually coffee and Mino, my coding companion.
Meet Mino