Hari Balakrishnan
Massachusetts Institute of Technology, MA·Computer Science
The cloud services and wireless networks that modern life runs on have become extraordinarily complicated — enormous in scale, built from many kinds of hardware, layered with software, and facing demands that change by the minute. Keeping them fast, reliable, and cheap is slow expert work, and the supply of people able to do it is not growing as fast as the complexity.
- Fields
Computer Science
Engineering
- Topics
networked systems
large language models
AI agents
- Grants
- 1 known grant, through Sep 2030
What they work on
- Test whether AI agents built on large language models can discover better decision rules, or heuristics, for networked systems in settings where goals are measurable, using concrete case studies and benchmark evaluation.
- Determine whether AI-discovered heuristics stay dependable when operating conditions shift at runtime, and whether they can adapt safely while a system is live.
- Develop a replay method that reconstructs the problem context of influential past systems research and asks whether an AI system independently proposes comparable abstractions and architectural concepts.
- Build evaluation methods that judge AI-generated system designs not only on performance but on understandability, robustness, and conceptual clarity, grounded in simulation, emulation, and hardware testbeds with expert review.
More about the research
This project asks a foundational question about recent artificial intelligence: can AI agents built on large language models (programs trained to generate and reason over text) genuinely take part in the kind of expert design work that today requires trained researchers and engineers? The goal is not to replace human judgment but to map where AI helps, where it fails, and why.
Why it matters
If AI agents can reliably assist with networked systems design, the cost, time, and scarce expertise needed to build the infrastructure behind commerce, communication, and public services could fall, while the risk of costly failures drops. The benchmarks, datasets, and evaluation tools produced give other researchers a shared way to test such claims, and findings about which parts of this expertise remain inherently human could shape how future engineers are trained.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Lingming Zhang
University of Illinois Urbana-Champaign, IL·Computer Science
Train LLMs jointly to generate synthetic software issues and to resolve them, using a self-play loop in which the model learns to create the issues that most improve its own bug-fixing ability.
Tianqi Chen
Carnegie Mellon University, PA·Computer Science·Early-career
Build a unified program representation that captures both high-level computation and data formats and low-level hardware optimizations of AI models, as the foundation for a new machine learning compiler.
Hae-Na Lee
Michigan State University, MI·Computer Science
Develop detection algorithms that analyze webpage layouts, interaction workflows, user actions, and accessible navigation paths to identify deceptive design patterns, including ones created by mismatches between a page's visual structure and its screen reader reading order.
The school
Massachusetts Institute of Technology: 25 labs on PI Pages, $1.1B a year on research, 692 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.