Arpit Gupta

Arpit Gupta

Assistant Professor in CS at UCSB

About me

I am an Assistant Professor in Computer Science at UC Santa Barbara, and Faculty Scientist at Berkeley Lab. I co-direct the Systems and Networking Lab (SNL) at UCSB.

At SNL, I have been utilizing my system-building skills to address a variety of pressing digital inequity challenges, namely, ensuring secure, performant, and affordable “Internet for All.” To this end, my current research focuses on democratizing the development of production-ready ML artifacts for self-driving networks (to ensure performant and secure connectivity with limited infrastructure and operational resources) and enabling data-driven policymaking (to ensure performant and affordable connectivity with limited capital resources).

Prospective Students

Join us in shaping a more equitable digital world! I am actively looking for Ph.D. students to join my group. In the next few years, our research group would extensively focus on developing network foundation models to further democratize the development of production-ready ML artifacts for self-driving networks and enabling data-driven policymaking. Please check out this invited talk that I recently gave at Monterey Data Conference’24 to get a gist of where we are headed as a research group.
Please find more details about my research here.

If you are interested, please reach out to me over email. I value diversity and inclusion in my research group and encourage applications from underrepresented groups. Also, it would help if you express genuine interest in the research problems that I am working on by reading some of our recent research papers.

Note: I am not an ML researcher, i.e., I do not make fundamental contributions to AI/ML algorithms that could be applied broadly to any application domain. I am a networked systems researcher who uses AI/ML to only solve networking problems.

Selected Publications

Please check this page for an extended list of publications.

Ongoing Projects

  • BQT: A tool that queries broadband plan offerings from major ISPs in the US at street-level granularity.
  • netFound: A foundation model for networking data that employs self-supervised learning techniques on abundant unlabeled network data, passively collected from production environment using PINOT for task-agnostic pre-training and smaller-scale labeled network data, actively collected using PINOT and netUnicorn for task-specific fine-tuning.
  • Trustee: A framework that cracks open decision-making for black-box ML models (for networks) using high-fidelity, low-complexity, and stable decision trees.
  • PINOT: A programmable data-collection infrastructure at UCSB to collect fine-grained (labeled) network data at scale.
  • netUnicorn: A data-collection platform that simplifies collecting network data for different learning problems from diverse network environments.

Workshops and Tutorials

As a junior researcher, it has been an absolute honor and privilege to get the opportunities to organize different workshops (and tutorials) on topics related to digital equity and self-driving networks.

News