Avatar

M Akash Kumar

Data & Applied Scientist

Microsoft

About Me

Hi, I’m Akash! I’m currently a Data & Applied Scientist in the Ads team at Microsoft.

I’m passionate about exploring new areas that can solve important real-world problems. I completed my B.Tech and M.Tech (Dual Degree) in Electrical Engineering from IIT Madras, graduating among the top 5% of students in my class. I worked under the guidance of Prof. Mitesh Khapra and Prof. B. Ravindran and published papers in top NLP conferences and journals such as TACL, ACL, and EMNLP. Previously, I interned with Microsoft, where my work on sponsored search led to a significant increase in revenue and user engagement in Bing and was published in KDD 2021.

Interests

  • Machine Learning
  • Natural Language Processing
  • Language Generation
  • Reinforcement Learning

Education

  • Dual Degree (B.Tech + M.Tech) in Electrical Engineering, 2021

    Indian Institute of Technology Madras

    CGPA: 9.50/10

Publications

Active Evaluation: Efficient NLG Evaluation with Few Pairwise Comparisons

We introduce algorithms to efficiently identify the top-ranked NLG system with few pairwise human annotations.
Akash Kumar M, Mitesh M. Khapra
March 2022 In ACL 2022
PDF

Diversity driven Query Rewriting in Search Advertising

We propose RL algorithms to generate diverse and high-quality query rewrites.
May 2021 In KDD 2021
PDF

A Survey of Evaluation Metrics Used for NLG Systems

We provide a detialed survey of various NLG evaluation metrics.
Ananya B. Sai, Akash Kumar M, Mitesh M. Khapra
April 2021 In ACM CSUR
PDF

Improving Dialog Evaluation with a Multi-reference Adversarial Dataset and Large Scale Pretraining

We propose a multi-reference adversarial dialogue evaluation dataset and an evaluation metric which is pretrained on a massive corpus …
Ananya B. Sai*, Akash Kumar M*, Siddhartha Arora, Mitesh M. Khapra
July 2020 In TACL 2020, MIT Press
PDF Code Website

Towards Transparent and Explainable Attention Models

We propose a modified LSTM cell with a diversity-driven training objective to improve the transparency and explainability of attention …
July 2020 In ACL 2020
PDF Code Video Slides
See all publications

Internships

 
 
 
 
 

Data & Applied Scientist Intern

Microsoft

May 2020 – Jul 2020
Developed large-scale multi-lingual language models for query rewriting in Bing Search Ads
 
 
 
 
 

Research Intern

Technical University of Darmstadt

May 2019 – Jul 2019 Darmstadt, Germany
Developed task-oriented chatbots that perform database transactions and query-related tasks in a conversational manner
 
 
 
 
 

Data Science Intern

LYNK Logistics

May 2018 – Jul 2018 Chennai, India
Developed an algorithm to suggest high demand locations to drivers in an e-hailing service by forecasting the spatio-temporal demand intensity maps
 
 
 
 
 

Deep Learning Intern

HyperVerge Inc.

May 2017 – Jul 2017 Bangalore, India
Developed a real-time Deep Learning based ATM surveillance system that was successfully demonstrated in several ATMs in Bangalore

Accomplish­ments

DAAD Scholarship

German Academic Exchange Service Jan 2019
Awarded the DAAD scholarship to pursue research internship in Natural Language Processing at German Universities

Kishore Vaigyanik Protsahan Yojana (KVPY) Fellowship

Department of Science and Technology, Government of India Mar 2016
Recipient of Kishore Vaigyanik Protsahan Yojana (KVPY) fellowship awarded by the government of India to exceptionally highly motivated students to pursue basic science courses and research career in science (ranked 273/1 lakh candidates).

Contact