Nikhil Kandukuri

MS CS (IIS) Graduate Student at Carnegie Mellon University

About Me

Hi, I’m a current graduate student at Carnegie Mellon University.

Possess two years of IT Experience working with the latest tech stack like Quarkus , gRPC , Google Pub/Aub , Debezium at Deutsche Bank.

Interned at MNCs like Samsung Research and Western Digital where I applied my skills of academia in some enriching projects centred around Bixby ( Samsung’s Voice Assistant) and the new generation of flash drives.

I am currently pursuing research in the space of event temporal extraction using LLMs. Have made a paper submission to NAACL 2024 as a co-author. Excited to explore further in the field of Mahcine Learning and Natural Language Processing!

Education

Carnegie Mellon University

Master of Science in Computer Science (IIS)

Aug 2023 - Dec 2024

BITS Pilani

Bachelors of Engineering in Electrical and Electronics Engineering

Aug 2017 - Jul 2021 (CGPA - 9.29/10)

Experience

Deutsche Bank

Senior Analyst

July 2021 - June 2023

• Played a key role in the development team that built the Cash Settlement Engine’ product for the Investment Banking Division

• The project’s aim was to replace existing five monolithic settlement systems with a modernistic cloud hosted micro-service based architecture acting as a unified settlement system across various products

• Cash Settlement Engine will save 5 million euros annually by decommissioning current vendor products

• Implemented key software strategies including Domain Driven Design, Event Driven Architecture, and Test Driven Development with code coverage exceeding 80%, and Behavior Driven Development with a 100% Passing Serenity BDD report

Samsung Research Institute

Associate

Jan 2021 - June 2021

• Worked in the Voice Intelligence R&D team focused on Automatic Speech Recognition.

• The outcome of the project was to enhance end to end Automatic Speech Recognition systems using data augmentation techniques.

• Worked with a variety of NLP frameworks related to speech recognition, synthesis, sampling, filtering and translation

Western Digital

Intern

May 2020 - July 2020

• At Western Digital I worked with the Firmware group in the RPG(Removable Product Group) department. The project was based in Data Analytics.

• The teams used .map files for observing patterns in the firmware. However, many crucial components were being looked at manually leading to inefficiencies in performance.

• The tool I developed parsed data from these files and converted them into a ubiquitous format, supporting SQL queries.

Projects

Neural Architecture Search for Skin Cancer Detection

Research Project UC San Diego

Enhanced diagnosis of early stages of skin cancer using the method of Neural Architecture Search (NAS), improved by the technique of Learning by Passing Tests (LPT) to improve accuracy by close to 1%

The experiments consisted of two models, a teacher model, and a learner model, mirroring processes occurring in human learning

Link to Preprint

Event Temporal Ordering using LLMs

Research Project Carnegie Mellon University

Directed benchmarking Large Language Models (LLMs) on event ordering temporal tasks for four existing datasets through prompt engineering and evaluated performance models encompassing LLaMa (7B,13B,70B) , Flan-T5(3B,XL,XXL)

Text Injection for Automatic Speech Recognition (ASR)

Research Project Carnegie Mellon University

Experimented bolstering ASR performance through representative alignment of text and speech upsampling text by a factor of 4 using contrastive loss a predominant technique in multimodal research

Awards and Acievements

Academic Merit Scholarship: Received scholarship for four semesters during my Undergraduate Degree, awarded based on GPA to the top performing 3% of the undergraduate cohort of close to 1100 students

Concours Romain Rolland: Third Rank All India under the B1 category (French Olympiad)

Deutsche Bank Global Hackathon: Declared as the best team from India for building an application promoting mental health

Deutsche Bank Recognition Award: Awarded the DBRecognition award for the first quarter of 2023 for migrating the Cash Settlement Engine from development to User Acceptance Testing (UAT).

Technical Skills and Courses

OS: Windows, Linux

Programming Languages: Python, Java, C++, C

Libraries: Gensim, OpenCV, TensorFLow, Kaldi, Espnet, Scipy, Scikit, Keras, Pytorch

Skills: CI/CD , Kubernetes, gRPC , Test Driven Development, SQL, PostgreSQL, Google Pub/Sub , Agile, Event Driven Architecture, Debezium

Subjects: Machine Learning, Artificial Intelligence, Neural Networks and Fuzzy Logic, Python Programming, Information Retrieval, Foundations of Data Science, Data Structures and Algorithms, C programming, Probability and Statistics, Operating Systems Object Oriented Programming, Digital Design, Optimization, Applied Statistical Methods, Generative AI, Multimodal Machine Learning, MLOps, Advanced Natural Language Processing, Speech Recognition and Understanding

Languages: English, Telugu, Hindi, French