Researcher at Fujitsu Research India

Ankush Agarwal

I work on knowledge-intensive NLP, large language models, question answering, and multimodal reasoning. My recent work focuses on making LLM systems more effective over structured knowledge, enterprise workflows, and complex documents.

ACL, NAACL, and EMNLP publications MS by Research, IIT Bombay Bengaluru, India
View Publications CV Google Scholar LinkedIn
131 Citations
6 h-index
6 i10-index
9 Publications

About

Research profile

I am a researcher at Fujitsu Research India, where I work on applied and foundational problems in language understanding. My work sits at the intersection of language models and structured knowledge, with an emphasis on question answering, retrieval and reasoning over tables and text, enterprise agent evaluation, and fine-grained multimodal document understanding.

I completed my MS by Research at the Indian Institute of Technology Bombay under the guidance of Prof. Pushpak Bhattacharyya. My academic and industry work shares a common theme: bringing reliable structure, retrieval, and reasoning into LLM-based systems.

Current focus

Knowledge-augmented LLM systems for enterprise and real-world reasoning tasks.

Recent venues

ACL 2024, NAACL 2025, ACL 2025, and EMNLP 2025.

Graduate work

MS by Research at IIT Bombay, CPI 8.79, advised by Prof. Pushpak Bhattacharyya.

Exam highlight

GATE 2020 score: 818, with 99.83 percentile.

Research

Core areas

Knowledge-infused language models

Methods for injecting structured and domain-specific knowledge into language models so they reason more accurately and with lower context overhead.

Question answering over mixed sources

Systems that combine text, tables, and knowledge graphs to answer questions that require aggregation, retrieval, and multi-hop reasoning.

Multimodal document understanding

Benchmarks and modeling approaches for finding fine-grained evidence in complex visual documents where localization and reasoning both matter.

Enterprise LLM evaluation

Evaluation environments for measuring how well LLM agents operate across fragmented data, access control boundaries, and real enterprise workflows.

Publications

Selected papers

Full publication list
2026

EnterpriseLab: A Full-Stack Platform for developing and deploying agents in Enterprises

Ankush Agarwal, Harsh Vishwakarma, Suraj Nagaje, Chaitanya Devaguptapu

arXiv preprint, March 23, 2026

Introduces a closed-loop platform for enterprise agent development that unifies tool integration, trajectory synthesis, training, and continuous evaluation.

Paper PDF
2025

Can LLMs Help You at Work? A Sandbox for Evaluating LLM Agents in Enterprise Environments

Harsh Vishwakarma*, Ankush Agarwal*, Ojas Patil, Chaitanya Devaguptapu, Mahesh Chandran

EMNLP 2025, Suzhou, China

Introduces EnterpriseBench, a 500-task benchmark for evaluating LLM agents in realistic enterprise settings with fragmented data and access controls.

Paper PDF
2025

Finding Needles in Images: Can Multi-modal LLMs Locate Fine Details?

Parth Thakkar*, Ankush Agarwal*, Prasad Kasu, Pulkit Bansal, Chaitanya Devaguptapu

ACL 2025, Vienna, Austria

Presents NiM-Benchmark and Spot-IT for evaluating and improving fine-grained detail localization in complex visual documents.

Paper PDF
2025

Hybrid Graphs for Table-and-Text based Question Answering using LLMs

Ankush Agarwal, Chaitanya Devaguptapu, Ganesh S

NAACL 2025, Albuquerque, New Mexico

Proposes a hybrid graph representation for question answering over mixed table-text sources, improving zero-shot results while reducing token usage.

Paper PDF
2024

HOLMES: Hyper-Relational Knowledge Graphs for Multi-hop Question Answering using LLMs

Pranoy Panda, Ankush Agarwal, Chaitanya Devaguptapu, Manohar Kaul, Prathosh A P

ACL 2024, Bangkok, Thailand

Builds distilled, query-aware knowledge graphs for multi-hop QA, improving answer quality while using substantially fewer tokens.

Paper PDF
2023

KITLM: Domain-Specific Knowledge InTegration into Language Models for Question Answering

Ankush Agarwal, Sakharam Gawade, Amar Prakash Azad, Pushpak Bhattacharyya

ICON 2023

Studies domain-specific knowledge infusion for question answering and shows gains over larger generic models in specialized settings.

Paper PDF
2022

Knowledge Graph - Deep Learning: A Case Study in Question Answering in Aviation Safety Domain

Ankush Agarwal, Raj Gite, Shreya Laddha, Pushpak Bhattacharyya, Satyanarayan Kar, Asif Ekbal, Prabhjit Thind, Rajesh Zele, Ravi Shankar

LREC 2022

Presents a knowledge graph guided QA system for aviation safety, grounded in accident reports and regulatory documents.

Paper PDF
2022

There is No Big Brother or Small Brother: Knowledge Infusion in Language Models for Link Prediction and Question Answering

Ankush Agarwal, Sakharam Gawade, Sachin Channabasavarajendra, Pushpak Bhattacharyya

ICON 2022

Explores knowledge infusion for both link prediction and question answering, connecting symbolic structure with language model behavior.

Paper PDF

* Equal contribution.

Experience

Resume highlights

2023 - Present Bengaluru, India

Researcher, Fujitsu Research India

Working on applied AI research spanning knowledge-graph-enhanced LLMs, question answering, enterprise agent evaluation, and multimodal document reasoning.

2021 - 2023 Mumbai, India

MS by Research, Indian Institute of Technology Bombay

Completed graduate research under Prof. Pushpak Bhattacharyya. CPI: 8.79. Program transition from MTech to MS by Research noted in LinkedIn profile.

Honors

Recent recognition

Nov 2025

FRIPL Award

Awarded for development of Needles in Images technology.

Aug 2025

Best Resource Paper Nomination

EMNLP 2025 nomination for work on enterprise LLM evaluation.

Jul 2025

SAC Highlights Award

ACL 2025 recognition for a top 1% submission.

Mar 2025

FRIPL Grand Award

Recognized for development of KG-RAG technology.

Contact

Get in touch

For research discussions, interviews, collaboration, or speaking opportunities, the fastest way to reach me is by email or LinkedIn.

Curriculum Vitae ankush98.12@gmail.com Google Scholar LinkedIn GitHub