Aishik Mandal

I am an ELLIS PhD student at UKP Lab, TU Darmstadt. I am supervised by Prof. Iryna Gurevych and co-supervised by Prof. Aurélien Bellet. I completed my Master's in Artificial Intelligence, Machine Learning and applications and my Bachelor's in Electronics and Electrical Communication Engineering at IIT Kharagpur. I also recieved the Silver Medal from the department of Electronics and Electrical Communication Engineering for the best academic performance among graduating students in the program.

At IIT Kharagpur, I worked at Centre of Excellence in AI with Prof. Jiaul Hoque Paik for my Master's Thesis. Prior to that, I was part of Complex Network Research Group and worked with Prof. Niloy Ganguly and Prof. Pawan Goyal, under whom I completed my Bachelor's Thesis. I did my summer internships at the Theoretical Foundations of AI group at Technical University of Munich under Prof. Debarghya Ghoshdastidar and at Articulabo, Cognitive Machine Learning Group at INRIA Paris under Prof. Justine Cassell. I recieved the DAAD WISE scholarship for my internship at TU Munich and the Charpak Lab scholarship and GKF International Internship scholarship for my internship at INRIA Paris.

Email  /  CV  /  LinkedIn  /  Google Scholar  /  Twitter  /  GitHub

profile photo

I'm interested in Dialogue Systems, Multimodality and Language Representation Learning. The projects I was part of are presented here.

Universal Transformer for Multimodal Self-supervised Learning
Aishik Mandal, Jiaul Hoque Paik
Master's Thesis, 2023

Developed a novel loss function to train an universal transformer encoder network for mapping both images and text in a common embedding space such that images and text conveying similar meaning are close to one another.

A Revenue Function for Comparison-Based Hierarchical Clustering
Aishik Mandal, Michael Perrot, Debarghya Ghoshdastidar
TMLR, 2023
codebase / paper[pdf]

Developed a revenue function to determine the goodness of a dendrogram obtained from hierarchical clustering algorithm in comparison-based settings.

Multimodal Turn Taking in Conversational Agents
Aishik Mandal, Biswesh Mahapatra, Justine Cassell
Summer Internship, 2022

Used multimodal data to improve turn taking of conversational agents in conversation with a human.

Discourse Mutual Information for Dialogue Understanding and Response Generation
Aishik Mandal, Bishal Santra, Pawan Goyal
Bachelor's Thesis, 2022

Developed a pretrained encoder for dialogues using a structure-aware Mutual Information based loss-function DMI (Discourse Mutual Information) and predicting response from only context encodings.

Representation Learning for Conversational Data using Discourse Mutual Information Maximization
Bishal Santra,Sumegh Roychowdhury,Aishik Mandal, Vasu Gurram,Atharva Naik,Manish Gupta,Pawan Goyal
NAACL, 2022
project page / paper[pdf]

Developed a pretrained encoder for dialogues using a structure-aware Mutual Information based loss-function DMI (Discourse Mutual Information).

Knowledge-Aware Neural Networks for Medical Forum Question Classification
Soumyadeep Roy,Sudip Chakraborty,Aishik Mandal,Gunjan Balde,Prakhar Sharma,Anandhavelu Natarajan,Megha Khosla,Shamik Sural,Niloy Ganguly
CKIM, 2021
codebase / paper[pdf]

Developed a knowledge aware BERT based model for improved medical text classification.

DAAD WISE Scholarship
Germany, 2021

Recieved DAAD WISE scholarship for summer internship at TU Munich

Charpak Lab Scholarship
France, 2022

Recieved Charpak Lab Scholarship for summer internship at INRIA Paris

GKF International Internship Scholarship
IIT Kharagpur, 2022

Recieved GKF International Internship Scholarship for summer internship at INRIA Paris

Extra Curricular Activities
National Service Scheme
IIT Kharagpur, 2018-2020

Volunteer of NSS and Co-Leader of the teaching team.

Student Welfare Group Mentor
IIT Kharagpur, 2020-2022

Mentor for 5 junior undergraduate students


Supporter of Liverpool FC and fantasy premier league player.


Love to read manga and watch anime.


Love to play open world games and Fifa.

Design and source code from Jon Barron's website