Dr. Arnab Phani
Dr. Arnab Phani
Postdoctoral Researcher
Affiliation: BIFOLD
Arnab is currently a postdoctoral researcher in the DEEM Lab research group headed by Sebastian Schelter at TU Berlin. He received his PhD from DAMS Lab at TU Berlin headed by Mathias Boehm. Prior to his PhD he was a Senior Software Engineer at Teradata Labs, Hyderabad, India.
Best Research Paper Award at EDBT 2025 for paper, "MEMPHIS: Holistic Lineage-based Reuse and Memory Management for Multi-backend ML Systems".
Systems for ML Data Management for Data Science Distributed Data Platforms Database Systems
Arnab Phani, Elias Strauss, Tim Schuster, Sebastian Schelter
ML Engineering Agents - A Data Management Perspective
Hao Chen, Arnab Phani, Sebastian Schelter
"Will This Data Break My Task?" - Interactive Synthesis of Task-Aware Data Unit Tests
Hao Chen, Arnab Phani, Sebastian Schelter
PrismaDV: Automated Task-Aware Data Unit Test Generation
Arnab Phani, Elias Strauss, Sebastian Schelter
stratum: A System Infrastructure for Massive Agent-Centric ML Workloads
BIFOLD contributes to CIKM 2026
At CIKM 2026, BIFOLD researcher Hao Chen will present PrismaDV, an AI system that generates data unit tests tailored to the code that later uses the data. The demo will be available on November 11, 2026.
Memphis paper receives 2026 ACM SIGMOD Research Highlight Award
The SIGMOD community honors the research of BIFOLD researchers Arnab Phani and Matthias Böhm. Their work on eliminating the inefficient reuse of intermediate computations across multi-backend machine learning systems addresses a key challenge at the intersection of data engineering and ML.
Researcher Spotlight: Dr. Arnab Phani
Dr. Arnab Phani is a Postdoctoral Researcher at BIFOLD, where he addresses data management challenges in modern AI. Building directly on his foundational PhD work at the DAMS Lab, his current research at the DEEM Lab focuses on enhancing runtime efficiency and fostering responsible data management practices across the entire machine learning pipeline - from data cleaning and validation to training and inference.