Vashu Chauhan

📍 India / Germany

📧 vashu22606@iiitd.ac.in

I am Vashu Chauhan, a final-year B.Tech Computer Science student at IIIT Delhi (GPA 8.72/10) with a deep interest in multimodal and language-based reasoning systems, interpretable feature discovery, and scalable data-centric pipelines for structured and unstructured learning tasks, as well as Multimodal Agentic Systems.

My research spans three exciting collaborations:

TU Darmstadt Research Fellow (Data and AI Labs, Feb 2026–Present) — investigating learning-driven optimization strategies for adaptive streaming data systems under real-time and resource constraints, advised by Prof. Manisha Luthra Agnihotri and Prof. Carsten Binnig.
Adobe Research Intern (Noida, Aug 2025–Present) — proposed FEST (Feature Engineering with Self-evolving Trees), a principled framework for automated, interpretable feature engineering; and architected Brand Genome, a knowledge-graph representation of visual brand identity.
IBM Research Intern (Remote, Aug 2024–Jun 2025) — designed a scalable data generation and validation pipeline for enterprise LLM fine-tuning; work nominated for Best Paper Award (Industry Track) at EMNLP 2025.

I have publications at EMNLP 2025, AAAI 2026, and papers under review at ICML 2026 and ECCV 2026. I was selected as a Research Fellow under the Athene Young Investigator Programme at TU Darmstadt and received a research travel scholarship ($1,750 USD) to present at the AAAI Student Abstract Track.

news

Feb 01, 2026	🎉 Selected as a Research Fellow under the Athene Young Investigator Programme at TU Darmstadt, working with Prof. Manisha Luthra Agnihotri and Prof. Carsten Binnig on adaptive streaming data systems.
Jan 15, 2026	🏆 Awarded a research travel scholarship ($1,750 USD) to present work at the AAAI 2026 Student Abstract Track.
Sep 10, 2025	📄 IBM Research work on enterprise LLM fine-tuning was nominated for Best Paper Award (Industry Track) at EMNLP 2025. Paper: Mind the Query: A Benchmark Dataset Towards Text2Cypher Task.

selected publications

EMNLP
Mind the Query: A Benchmark Dataset Towards Text2Cypher Task

Vashu Chauhan and others

In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025

Abs Bib

We introduce a benchmark dataset and evaluation framework for the Text2Cypher task, enabling natural language interfaces over graph databases. The work was nominated for Best Paper Award (Industry Track) at EMNLP 2025.
@inproceedings{chauhan2025mindquery, title = {Mind the Query: A Benchmark Dataset Towards Text2Cypher Task}, author = {Chauhan, Vashu and others}, booktitle = {Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing}, year = {2025}, }
AAAI
IMPACT: Integrated Multimodal Pipeline for Accident Tracking

Vashu Chauhan and others

In Proceedings of the AAAI Conference on Artificial Intelligence (Student Abstract), 2026

Abs Bib

A multimodal pipeline that combines vision and language models for real-time traffic accident tracking and situational awareness.
@inproceedings{chauhan2026impact, title = {IMPACT: Integrated Multimodal Pipeline for Accident Tracking}, author = {Chauhan, Vashu and others}, booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence (Student Abstract)}, year = {2026}, }
ICML
Scaling Controllable Modeling via Self-Evolving Feature Engineering

Varun Khurana, Vashu Chauhan, and others

International Conference on Machine Learning (Under Review), 2026

Abs Bib

Proposes FEST (Feature Engineering with Self-evolving Trees), a principled framework for automated, interpretable feature engineering that recovers expert-level feature coverage on structured and unstructured data.
@article{khurana2026fest, title = {Scaling Controllable Modeling via Self-Evolving Feature Engineering}, author = {Khurana, Varun and Chauhan, Vashu and others}, journal = {International Conference on Machine Learning (Under Review)}, year = {2026}, }
ECCV
Traffic Accident Causality Reasoning

Vashu Chauhan and others

European Conference on Computer Vision (Under Review), 2026

Abs Bib

Addresses causal reasoning for traffic accident analysis using multimodal evidence, combining structured scene understanding with language-based causal inference.
@article{chauhan2026eccv, title = {Traffic Accident Causality Reasoning}, author = {Chauhan, Vashu and others}, journal = {European Conference on Computer Vision (Under Review)}, year = {2026}, }