logo True scholar network
Back to Top

About

Arnab Kar is a multifaceted data scientist and machine learning expert with a rich blend of academic, entrepreneurial, and corporate experience. Currently pursuing a Ph.D. in Computer Science at Duke University, Arnab specializes in cutting-edge AI and ML technologies, with a strong focus on applications in fintech, real estate, and misinformation detection. His academic journey is backed by a solid foundation in electrical engineering and computer science, and he has consistently applied this knowledge to innovative projects that intersect technology and business strategy. Arnab's entrepreneurial spirit has led him to co-found startups such as Mitify+ and EnergyComply Tech, where he developed intellectual property and technology stacks for misinformation detection and real estate compliance tools, respectively. At Mitify+, he pioneered solutions for misinformation monitoring, while at Exchange Robotics, he contributed to the development of unique technological solutions for the illiquid credit market, engaging with major stakeholders like Moody’s and S&P to enhance platform capabilities. Arnab's corporate experience is equally impressive, having held key roles at Saks Fifth Avenue, where he implemented advanced machine learning models for demand forecasting and improved inventory management processes. His time at E Ink Corporation saw him lead cross-functional teams in developing simulations for ink display behavior, contributing to the company’s patent portfolio and significantly reducing product development timelines. Arnab’s passion for innovation extends beyond his professional roles—he is an active contributor to sustainability initiatives and mentorship programs, aiming to inspire the next generation of tech leaders. His diverse skill set encompasses computational finance, natural language processing (NLP), operations research, and advanced analytical modeling. Throughout his career, Arnab has demonstrated an exceptional ability to align technological advancements with strategic business goals, making him a sought-after leader and innovator in the AI and data science community.

View More >>

Skills

Experience

placeholder
Lead Data Scientist

MITIFY+

Apr-2023 to Present
placeholder
Co-founder, ML and Data Engineering AI

Exchange Robotics (ExR)

May-2023 to Nov-2023
placeholder
Deep Learning Engineer II

E Ink Corporation (Eink)

Feb-2021 to May-2022
placeholder
Machine Learning Scientist

Saks Fifth Avenue, US

Jun-2022 to Feb-2023
placeholder
Senior Data Scientist

Sirion

May-2020 to Feb-2021
placeholder
Student Research Intern

Indian Statistical Institute, Kolkata

May-2015 to Dec-2015
placeholder
Research Intern

IIT Kharagpur (Indian Institute of Technology)

May-2014 to Dec-2014
placeholder
Student Research Associate

IIT Kanpur (Indian Institute of Technology, Kanpur)

May-2017 to Aug-2018
placeholder
Pre Doctoral Intern

Institute of Science and Technology Austria (ISTA)

Jan-2018 to Aug-2018

Education

placeholder
Indian Institute of Information Technology Allahab...

B.Tech in Information Technology(IT)

Passout Year: 2025
placeholder
Duke University (DU)

Ph.D. in Computer Science

Passout Year: 2018

Peer-Reviewed Articles

Internal and External Re-keying and the way forward

Side Channel Analysis are the security attacks due to the issues in the implementations. This attack bypasses the mathematical security provided by the cryptographic algorithms. These attacks are broadly categorized into the issues related to architectural of the chip manufacturing, attack due to unwanted leakages like power leakage, acoustic leakage, thermal leakage or electromagnetic leakages, and the issues due to programming vulnerabilities for example the heartbleed bug etc. The architectural related issues are fixed when the newer version of hardware is designed once the vulnerability is found in the earlier version. The programming related attacks are solved by patching the software and updating the code that caused the vulnerability to be exploited. The leakage issues are the ongoing issues since it was first discovered in 1997. Among the various leakage issues, the acoustic and thermal leakages aids in the attack related to power analysis. The Electromagnetic attack boils down to the power analysis issue and hence, it all comes down to the power analysis attack. Since it was discovered, the researchers have suggested the solutions for them but on the other side, they would also be vulnerable again. The Power analysis attacks are mainly classified into Simple Power Analysis (SPA), Differential Power Analysis (DPA), Correlation Power Analysis (CPA), and profiled attacks. Their countermeasures are mainly masking and rekeying apart from architectural changes. The masking has been researched extensively and have been widely implemented countermeasure. However, it comes with a very big overhead. Therefore, the researchers started exploring the rekeying to counter them. Rekeying has been classified mainly into the internal and external rekeying both having its advantages and disadvantages. There is currently no literature available that discusses both in detail. This work surveys the work on both the approaches and suggest the way forward for the researchers of the re-keying.

DEVELOPING A DATA-DRIVEN ARCHITECTURE FOR IMPLEMENTING AI-ENABLED DYNAMIC PRICING STRATEGIES IN THE AUTOMOTIVE INDUSTRY

In the Automotive Industry, dynamic pricing is used a lot to make the most money and hold off the competition. The Automotive industry is using AI to build a data-centric framework that will allow dynamic pricing. This research will look at how they are doing it. Automakers can find out about how customers act, how the market is changing, and how competitors plan to beat them by using complicated formulas and strict data collection methods. The aim of this research is to analyze how dynamic pricing protects prices in various industries, with a particular focus on its application in the automotive industry. In addition, the research will discuss about data-driven design approaches incorporating with artificial intelligence (AI), mainly how these technologies could be used to improve pricing strategies by automating choices and letting prices adjust based on the market. Important things like how to use market trends to our advantage, gather and analyze data, and understand how customers behave, and merchandise sales are the focus areas of the paper. As part of the project, AI could also be used to improve pricing methods. Some of these are prediction analytics, machine learning, and reinforcement learning. We can figure out how to make the most money and guess what prices will be in the future by using algorithms that look at past price data. Finally, the study shows that price strategies that are driven by AI and design that is driven by data can have a big impact on the automotive industry. Businesses in the Automotive industry might be able to boost competition, new ideas, and customer trust by using dynamic pricing systems and staying honest all the way through.

Scalable Data Partitioning and Shuffling Algorithms for Distributed Processing: A Review

Scalable data splitting and shuffle algorithms have emerged as crucial elements of effective data processing in distributed computing and big data. This article provides an in-depth analysis of the complex terrain of these algorithms, which play a crucial role in ensuring efficient data distribution, load balancing, and resource optimisation in distributed systems. Among the most important discoveries are the varying functions performed by algorithms like hash-based, range-based, and sort-based techniques. The importance of measurements like data transmission overhead, processing time, and network utilisation in illustrating the impact of various algorithms on performance is emphasised. Challenges, such as algorithmic complexity and the never-ending search for efficiency and adaptation, remain despite their evident importance. The ramifications affect a wide variety of parties. Adaptive algorithms, privacy protection, and energy efficiency are all areas where researchers may make strides forward. Insights for optimised data processing operations, including careful algorithm selection and performance adjustment, might benefit practitioners. Leaders are urged to appreciate the algorithms' strategic value in realising data-driven goals and to invest wisely in the systems and personnel needed for effective distributed processing. As a result, organisations are able to extract meaningful insights, make informed real-time decisions, and navigate the ever-changing world of big data to scalable data division and shuffling algorithms.

Projects

May-2017 to Oct-2018

Multi-agent Generative modeling

Generative Modeling to be able to *understand-ably* generate images (or scenarios). It could reason about the entities (people, things, objects, background) in generated image (and between images). Also applicable to other use-cases such as: ☆ Financial use-case: macro-economic scenario modeling to understand market dynamics with different *risk-leveled* agents/participants (macro-economic scenarios being one of the agent behavior). ☆ Business scenario modeling: impact on different lines of business, and bottom-line under different competitive dynamics, macro-economic conditions, and regulatory environments. ☆ Insurance use-cases: modeling the impact of different market conditions on different industry sectors/sub-sectors (and corresponding lines of underwriting applications).
...see more
Aug-2019 to May-2020

Time (processor) and Space (memory) optimized Machine Learning hardware-accelerator scheduler

Designed, implemented, and tested scheduling techniques to optimally use hardware resources for Machine Learning payloads on tabular-databases (SQL queries, for example), while being under time-limits. Also applicable for following kinds of use-cases: ☆ Project Management: Time and resource optimization for complex engineering/supply-chain/industrial projects, with embedded risk management and contingency planning. ☆ Fleet management: using optimal resources (human resources, and capital), while respecting client requirements, and tolerance limits.
...see more
Sep-2015 to Jun-2016

Provably-verified Data-security when performing data analysis on permission-ed data

Used mathematical techniques to ensure data analysis on dis-aggregated data respects ownership and access-control rules set on data sources. Other use-cases could be:: ☆ Ensuring AI generated content, or AI-enabled knowledge discovery respects ownership rules of underlying data-source; not revealing confidential data. ☆ Ensure financial or corporate reporting does not leak business logic or confidential information: revealing information at the right level of granularity between organizational silo-s, up the organizational pyramid, and/or in public disclosures.
...see more
Aug-2014 to Jun-2015

Knowledge-graph based fact-checks and relationship-discovery in Language-Model generated content

Using graph queries (on knowledge graph) to fact-check content, and verify logical claims (which might not immediately be in the knowledge graph). Other applicable use-cases could be:: ☆ Legal-tech use-case: Enable deductions and implications in corporate and legal communications, enabling better dispute handling. ☆ Insurance verification: claim checks, fact verification, policy-alignment checks. Tool would enable improved cash-flow, reduced administrative burden, improved transparency.
...see more
Jan-2018 to Aug-2018

Distributed Data-Influence detection for (any) Machine Learning model (Interpretability, explainabil...

Developed methods that could explain model behavior, at scale, as it relates to underlying training data. Other similar use-cases could be: ☆ ML Trust and Safety use-cases: ensuring trust in models by attributing model behavior to data; de-biasing models from gendered/societal/cultural artifacts. ☆ De-biasing model-enabled decision-making on protected attributes (gender, race, etc).
...see more

Certificates

Issued : Feb 2025
  • dott image By : Duke University
  • dott image Event : Computational M...
Computational Microeconomics

Scholar9 Profile ID

S9-112024-1206394

Publication
Publication

(0)

Review Request
Article Reviewed

(19)

Citations
Citations

(0)

Network
Network

(0)

Conferences
Conferences/Seminar

(0)

Academic Identity

SSRN ID
SSRN ID