Postdoctoral Researcher

Debasmita Lohar

“The true measure of success is how many times you can bounce back from failure.” — Stephen Richards

Email: debasmita.lohar [at] kit [dot] edu
Address: Room 202, Building 50.34,
KIT / Institute of Information Security and Dependability (KASTEL)
Debasmita Lohar

About Me

Hi! I am a Postdoctoral Researcher working with Bernhard Beckert in the Application-oriented Formal Verification group at KASTEL — Institute of Information Security and Dependability, Karlsruhe Institute of Technology.

I obtained my PhD from Saarland University in collaboration with the Max Planck Institute for Software Systems (MPI-SWS) Saarbrücken, where I was advised by Eva Darulova. Before that, I was a graduate student at the Indian Institute of Technology (IIT) Kharagpur, advised by Soumyajit Dey.

My primary research areas are: Program Analysis, Approximate Computing, Probabilistic Analysis, with applications in embedded systems, scientific computing, and machine learning.

News

Paper Our work, "Cost of Soundness in Mixed‑Precision Tuning" is accepted at OOPSLA 2025.
Paper Our work, "Of Good Demons and Bad Angels: Guaranteeing Safe Control under Finite Precision" is accepted at FMCAD 2025.
Call I have a master's thesis project — see topics.

Recent Publications

Cost of Soundness in Mixed-Precision Tuning

Anastasia Isychev and Debasmita Lohar — accepted at OOPSLA 2025.

Preliminary Version Code

Of Good Demons and Bad Angels: Guaranteeing Safe Control under Finite Precision

Samuel Teuber, Debasmita Lohar, and Bernhard Beckert — accepted at FMCAD 2025.

Extended Version

Towards Precision-Aware Safe Neural-Controlled Cyber-Physical Systems

Harikishan Thevendhriya, Sumana Ghosh, and Debasmita Lohar — ESWEEK-ESL 2024.

Preprint Presentation Tool: Nexus

Sound Mixed Fixed-Point Quantization of Neural Networks

Debasmita Lohar, Clothilde Jeangoudoux, Anastasia Volkova, and Eva Darulova — ESWEEK-TECS 2023.

Paper Presentation Tool: Aster

A Two-Phase Approach for Conditional Floating-Point Verification

Debasmita Lohar, Clothilde Jeangoudoux, Joshua Sobel, Eva Darulova, and Maria Christakis — TACAS 2021.

Paper Talk Tool: Blossom

Sound Probabilistic Numerical Error Analysis

Debasmita Lohar, Milos Prokop, and Eva Darulova — iFM 2019.

Paper Presentation Tool

View all on Google Scholar

Service

Teaching

Karlsruhe Institute of Technology

Saarland University

Indian Institute of Technology Kharagpur

  • Fault Tolerant Systems (Spring 2016, 2015, 2014)
  • Theory of Computation (Fall 2015)
  • Computer Organization and Architecture Lab (Fall 2014)

Honors and Awards

Theses