Rishab Vempati
I'm a Computer Science & Artificial Intelligence undergraduate at Purdue University. I research computer-use agents in the CIDER Lab with Prof. Jason Wu, specifically training vision-language models to navigate and evaluate user interfaces.
More broadly, I'm interested in machine learning beyond the agent setting: how models are trained and post-trained, reinforcement learning, and building the systems that make them work in practice. I'm always open to research collaborations and new problems in these areas.
News
- Aug 2026Joined IBM watsonx Orchestrate as a Software Engineering Intern.
- Feb 2026Started research in the CIDER Lab with Prof. Jason Wu; released uxCUA as an arXiv preprint.
Off the clock, I'm usually on a golf course — I track my game with a handicap tracker I built.
Research
Agentic systems, Reinforcement Learning for VLMs, and Computer Vision
Computer-Use Agents for UI Usability
Feb 2026 – Present- Co-authored uxCUA, training VLM computer-use agents with reinforcement learning at scale.
- Built distributed evaluation infrastructure executing 21K+ computer-use-agent rollouts across 879 websites on H100s.
- Built Python pipelines for large-scale LLM/VLM evaluation across model variants and task families, tracing success, latency, failures, and comparative performance on SLURM-managed GPUs.
Probabilistic Demand Forecasting
Jan 2024 – May 2024- Built Bayesian model-selection pipelines across 10K+ SKUs, automating demand-forecast evaluation at scale.
- Built Isolation Forest and Gaussian Mixture anomaly-detection pipelines for demand time series and integrated them into forecasting workflows, improving downstream SKU-level forecast accuracy by 3%.
Parking Lot Detection & Mapping
2024- Developed a U-Net segmentation model for satellite imagery in a 5-person team, achieving an 84% IoU accuracy.
- Built an automated pipeline ingesting 50GB+ of Bing Maps imagery and containerized the training workflow with Docker.
- Contributed 5,000+ validated parking-lot annotations to OpenStreetMap for open-source urban mapping.
Publications
- [1]
Training Computer Use Agents to Assess the Usability of Graphical User Interfaces
Alice Gao, Weixi Tong, Rishab Vempati, Katharina Reinecke, R. Benjamin Shapiro, Tianyi Zhang, Jason Wu.
arXiv preprint arXiv:2604.26020.
- [2]
Enhancing Ensemble Algorithm & Calibration Process for Demand Forecasting
Aman Kumar, Arhan Chaudhary, Kapil Naresh, Rishab Vempati, Sahil Desai, Shreya Gupta, Siddh Patani, Siddharth Sancheti, Zachary Kirkeby.
The Data Mine, Purdue University, 2024.
Experience
Engineering roles building production AI and distributed systems.
Software Engineering Intern · IBM — watsonx Orchestrate
Container image optimization and AI-agent tooling.
Founding Engineer · NovusMinds AI
Built Cora, an enterprise knowledge platform.
Software Engineer, Part-Time · Ekai
RAG + MCP retrieval for enterprise knowledge.
Software Engineering Intern · AssetMark
Event-driven microservices on Azure.
Software Engineering Intern · Accelera Payments
Kafka streaming for ISO 20022 payments.
Get In Touch
I'm open to connecting about research collaborations, graduate school, or interesting problems in agentic AI.
Say Hello