Perception Flight deck
Inverse-LLaVA
Rethinking how vision and language meet, with no alignment pre-training and 45% fewer training samples.
Read case study
Research × Engineering
From research toworking AI.
I turn advances in multimodal learning and recommendation into systems people can use.
Scroll to unfold four projects.
Selected work
The work, unpacked.
Four projects, from multimodal learning to production infrastructure.
People Crew cabin
Treverse Recsys
Personalized recommendations, from temporal graph learning to real-time serving and controlled experiments.
Read case study
Dialogue Service bay
AI for Negotiation
Turning negotiation transcripts into research data, reducing annotation cost from $5,000 to $3 per transcript.
Read case study
Delivery Propulsion assembly
FORGE
A shared ML platform connecting training, deployment, and observability across cloud and edge workloads.
Read case study
02 / 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.
01 / 05
Open project
Frame the question.
Ancient Mortars structures roughly 10 million particle images for comparison.
02 / 05
Open project
Fit the representation.
Inverse-LLaVA maps text into visual space inside the language model.
03 / 05
Open project
Make the comparison legible.
Negotiation transcript coding reports a cost drop from more than $5,000 to $3.
04 / 05
Open project
Publish a reviewable release.
FORGE defines six contracts across distinct ML runtimes.
05 / 05
Open project
Return field evidence.
Treverse joins controlled exposure with reproducible learning and serving.
03 / Experience