Professor of Computer Science · University of Florida

Sumit Kumar Jha

Formal methods for trustworthy AI, from model checking at Carnegie Mellon to controlling agentic AI today.

Verification and synthesis techniques that make learned systems interpretable, controllable, and certifiable. U.S. citizen.

$17M+ in sponsored research as PI across 16 federal programs
Published at NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, AAAI, IJCAI, DAC, and ICCAD
Portrait of Sumit Kumar Jha
’04
B.Tech. (Honors), CSE
IIT Kharagpur
’09
M.S., Computer Science
Carnegie Mellon University
’10
Ph.D., Computer Science
Carnegie Mellon University
Now
Professor, Computer Science
University of Florida
About
2005–12Model checking of hybrid and biological systems at Carnegie Mellon with Ed Clarke, Bruce Krogh, and Chris Langmead; internships at Microsoft Research and GM R&D.
2013–19Neuro-symbolic AI and formal synthesis for in-memory computing; AFOSR Young Investigator; four AFRL visiting appointments.
2020–Formal methods for trustworthy and agentic AI, AI for science: DARPA GARD, TIAMAT, CLARA, and UPTAIC; ONR Science of AI; DOE autonomous labs.

Sumit Kumar Jha works on formal methods and trustworthy AI: verification and synthesis techniques that make learned systems interpretable, controllable, and certifiable. One line of work holds machine learning to logical specifications: counterexample-guided synthesis with LLMs, temporal-logic grounding of natural language, formal verification of chain-of-thought faithfulness, and contract enforcement over multi-agent trajectories. A second develops attribution and steering methods for inspecting and controlling model internals. Both now apply to AI for science: scientific foundation models and the formal analysis of their internal representations.

He has served as lead or prime PI on DARPA CLARA, GARD, TIAMAT, and UPTAIC awards, NSF SPX and SHF awards, ONR Science of AI, and a DOE ASCR multi-institution project, with additional support from AFRL/AFOSR and national laboratories. His work appears at AAAI, ACL, CVPR, DAC, EMNLP, ICCAD, ICLR, ICML, IJCAI, and NeurIPS.

Research

Four areas of current work
{{ p.num }}

{{ p.title }}

{{ p.desc }}

Supported by
{{ sp.agency }}{{ sp.program }}
{{ p.paper }} · {{ p.venue }}

Featured papers

All 173 publications →
{{ f.figAlt }}
{{ f.venue }} {{ f.badge }}

{{ f.title }}

{{ f.blurb }}

{{ s.t }}
Cover of AI for Cybersecurity: Research and Practice
Book · Wiley / IEEE Press · 2026
AI for Cybersecurity: Research and Practice
H.H. Song, E. Bertino, A. Velasquez, H.H. Wang, Y. Shoshitaishvili, S.K. Jha
Funding
$17M+

Sponsored research led as PI, across sixteen federal programs from DARPA, NSF, DOE, ONR, and the Air Force, including a $5.1M six-institution DOE consortium with Argonne and a $4.5M DARPA prime award.

{{ s.name }}
{{ s.full }}
{{ g.program }} · {{ g.name }}
{{ g.desc }}
{{ g.role }}

Students

Sumit Kumar Jha with his research group at the University of Florida
The group at Malachowsky Hall, University of Florida, 2026.
11
Ph.D. graduates
5
Postdocs mentored
6
M.S. theses
33
Undergraduate researchers
Current Ph.D. students · University of Florida
{{ st.name }}{{ st.from }}

Alumni have gone on to faculty positions (CU Boulder, Oakland, UCF), national laboratories (AFRL, ORNL), government, and industry (Google, Lockheed Martin, Sanofi Pasteur).

Work with me

Collaboration, advising, and speaking.

Twenty years of turning specifications into guarantees: for hybrid systems, for in-memory hardware, and now for foundation-model agents. Available for engagements.

Start a conversation
{{ o.title }}
{{ o.desc }}
Collaborators & partners
{{ pt }}
Patents & technology transfer
{{ pa.no }} · {{ pa.title }} ({{ pa.assignee }})

Talks & press

Invited talks & panels
{{ tk.year }}
{{ tk.title }}
{{ tk.venue }}

News

Honors
{{ h }}
{{ n.date }}

{{ n.text }}