EDUCATION

University of Massachusetts Amherst

Expected Graduation: May 2025
Master of Science Computer Science
GPA: 3.89/4

Birla Institute of Technology and Science - Pilani (BITS Pilani)

August 2018 – June 2023
B.E. Computer Science and M.Sc. Mathematics (Dual Major)
GPA: 9.15/10 (Graduated with distinction)

Coursework

Data Science Courses: Foundations of Data Science, Optimization, Applied Statistical Methods, Statistical Inference and Applications, Advanced Machine Learning (ML), Information Retrieval, Advanced Natural Language Processing (NLP), Reinforcement Learning, Intelligent Visual Computing. Other Courses: Database Systems, Data Structures and Algorithms, Object-Oriented Programming, Compiler Construction, Computer Architecture, Computer Networks.

EXPERIENCE

Prompt Specificity - Unlocking Secrets of LLM Behavior (GitHub)

Amazon Research (January 2024 – August 2024) Supervisors: Prof. Andrew McCallum, Dr. Haw-Shiuan Chang

  • Conducted an in-depth analysis on the effect of prompt specificity on the performance of Large Language Models (LLMs) in long-form text generation, developing a novel benchmark dataset for evaluating LLM-generated narratives. [Accepted at WNU workshop, EMNLP]
  • Designed and implemented an evaluation benchmark to assess the trade-off between instruction satisfaction and long-form text generation quality, noting over 30% depreciation in LLM performance due to the trade-off.
  • Analyzed the impact of prompt specificity on 7B models to understand training data properties and their effects on different stages of model training, including supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF). [Published]

Database Migration

Database Administrator at First Meridian - OnSolve LLC (January 2023 – June 2023 in Bangalore, India)

  • Managed database geocoding and migration from local onsite servers to cloud platforms.
  • Built Terraform scripts to monitor various database activities/thresholds on AWS such as database shutdown, CPU utilization, database storage limits, etc.

Hidden and Face-like Object Detection - Computer Vision (GitHub)

Undergraduate Thesis under Prof. Subrata Chakraborty at the University of New England (August 2022 – December 2022 in Sydney, Australia)

  • Developed CNN-based classifiers and detectors to differentiate faces from face-like objects and identify face-like objects hidden amidst background clutter with an accuracy of over 80% in different settings. [Published]
  • Conducted psychological image, left-brain right-brain, and Rorschach inkblot tests on image classifiers to identify if they possess biases towards any particular class of objects and mimic human personality traits. [Published]

Subject Line Generation using Artificial Intelligence and NLP (GitHub)

Data Science Intern at Epsilon India (January 2022 – April 2022 in Bangalore, India)

  • Built a transformer model to generate diverse email subject lines given a theme or starting word(s) as the input.
  • Generated 10 semantically similar subject lines by considering it an English-English neural machine translation task.
  • Built an LSTM model to predict open rates of emails that use the generated subject lines with a 91% accuracy.

Travel Network Optimization using Pyomo

Summer Research Intern at Ecom Express Pvt. Ltd. (May 2020 – July 2020 in New Delhi, India)

  • Investigated and modeled different logistics problems including the Travelling Salesman, Capacitated Vehicle Routing, and Pick-up Delivery Problems using Pyomo.
  • Built a Pyomo-based Python module to generate the shortest Hamiltonian cycle connecting ‘n’ delivery locations.

DATA SCIENCE PROJECTS

Impact of LLM Quantization on Text Quality

Supervisor: Prof. Mohit Iyyer (August 2024 – Present)

  • Conducting experiments to study the impact of quantization and speculative decoding on the quality of LLM-generated text, particularly instruction-following, factuality, and hallucination.

Cross-domain Personalization for LLMs

Supervisor: Prof. Hamed Zamani (May 2024 – Present)

  • Creating a benchmark dataset for fine-tuning LLMs to generate personalized responses based on data from multiple sources such as Google Scholar, LinkedIn, Twitter, and home pages.

BioNLP - Self-rewarding LLMs for Medical Question Answering

Supervisor: Prof. Hong Yu (May 2024 – August 2024)

  • Implemented the self-rewarding pipeline using Biomistral-7B, tailored for medical question-answering systems, to enhance AI responsiveness and accuracy in healthcare applications.
  • Created synthetic evaluation fine-tuning data using customized LLM-as-a-Judge prompts.

BioVision - Deep Learning for Medical Image Analysis (GitHub)

Research Assistant under Prof. Subrata Chakraborty at the University of New England (May 2024 – August 2024)

  • Created a transfer learning framework to classify non-verbal vocalizations from minimally speaking autistic individuals by considering audio data as images using Mel spectrograms. Achieved a state-of-the-art F1 score of 0.79, compared to the previous best of 0.52. [Under Review]
  • Conducted systematic literature reviews on the applications of deep learning in radiology for lung cancer and fibromyalgia syndrome diagnostic tasks including detection, segmentation, classification, and therapy outcome prediction. [Published]

Dynamic Fine-Tuning of LLMs Using RLHF for Improved Alignment (GitHub)

Course Instructor: Prof. Mohit Iyyer (February 2024 – May 2024)

  • Presented a proof-of-concept to enhance LLM alignment with human values through dynamic prompt generation based on intermediate model performance, resulting in improved model accuracy, optimization, and alignment.
  • Introduced and evaluated two novel training approaches—Always-On-Policy and Starts-On-Policy—with generations being preferred 61.85% of the time compared to traditional RLHF training. [Under Review]

Adaptive Learning - Classification of Math Word Problems Based on Difficulty Level (GitHub)

January 2022 – April 2022

  • Built an ensemble learning classifier integrating features from word embeddings, parts of speech tags, and readability scores to predict the difficulty level of a given math word problem. Achieved an AUC score of 0.92.

Information Retrieval - Course Project (GitHub)

January 2022 – April 2022

  • Developed a Boolean information retrieval system capable of processing wildcard queries and correcting spelling errors using the Permuterm method and edit distance, respectively.
  • Implemented the PageRank and HITS algorithms from scratch to compute the static and dynamic page ranks, respectively, based on a given web graph and query.

Emotionally Aware Conversational Agent for Mental Health (GitHub)

August 2021 – December 2021

  • Built a multi-faceted emotionally aware chatbot combining rule-based, retrieval-based, and generative methodologies.
  • Trained LSTMs and GRUs for the retrieval-based module and built a seq2seq encoder-decoder-based text generator for the generative module.

Prediction of Video Game Development Problems Using Word Embeddings (GitHub)

July 2021 – October 2021

  • Built machine learning, ensemble learning, and multi-layer perceptron models, achieving an AUC score of 0.97, to classify video game development problems into different groups based on the quote and problem description. [Published]
  • Performed a comparative study based on AUC scores and Friedman’s mean rank test by implementing five word embedding techniques, SMOTE, three feature selection techniques, five ML classifiers, and three ensemble learners. [Published]

OTHER PROJECTS

Query Optimization for Relational Database Management Systems - Study Project

June 2022 – September 2022

  • Performed a review of join query, semantic query, and hybrid query optimization models that use deep-reinforcement learning.

Compiler Construction - Course Project (GitHub)

January 2022 – April 2022

  • Designed a miniature programming language and developed a DFA-based lexical analyzer in C++ from scratch without using Lex.
  • Built a parser that takes in lexer-generated tokens as the input and reports errors in the code. Built an intermediate code generator to output intermediate code as quadruples for program construct expressions and conditional statements.

Database and Inventory Management for CCIT at BITS Pilani, Hyderabad Campus

August 2021 – December 2021

  • Built a robust web application using HTML, CSS, and JavaScript to aid the CCIT department in managing inventory. The application was developed with the ER model of the relational database in Boyce-Codd Normal Form (BCNF).
  • Provided features for creating new user accounts, querying and searching using an advanced search and filter bar, printing summary sheets, sending inventory-related requests, collecting feedback, and having multiple views for different kinds of users.

Travel and Hotel Booking - Web Application (GitHub)

January 2021 – April 2021

  • Built a full-stack web application using HTML, CSS, and Flask for travel and hotel booking.
  • The MySQL-based database was developed with the ER Model entirely in BCNF form.

Track and Trigger - Android App Development

August 2020 – November 2020

  • Built a Java-based Android application on Android Studio and Firebase to help people manage their daily home inventories and keep track of day-to-day activities using alarms and reminders.
  • Implemented additional features like Gmail/Facebook login, image and note sharing on WhatsApp, and customizable to-do lists.

Mathematical Modelling of COVID-19

February 2020 – May 2020

  • Worked on mathematically modeling the spread, peak, and reduction of the COVID-19 pandemic. [Published]

PUBLICATIONS

  • Anirudh Atmakuru, Subrata Chakraborty, Massimo Salvi, Oliver Faust, Prabal Datta Barua, Makiko Kobayashi, Ru San Tan, Filippo Molinari, Abdul Hafeez-Baig, U Rajendra Acharya, “Fibromyalgia detection and diagnosis: A systematic review of data-driven approaches and clinical implications (2013-2023)”, IEEE Access (2025). [Link]

  • Anirudh Atmakuru*, Jatin Nainani*, Rohith Siddhartha Reddy Bheemreddy*, Anirudh Lakkaraju*, Zonghai Yao, Hamed Zamani, Haw-Shiuan Chang*, “CS4: Measuring the Creativity of Large Language Models Automatically by Controlling the Number of Story-Writing Constraints”, In 6th Workshop on Narrative Understanding (WNU) at Empirical Methods in Natural Language Processing (EMNLP), 2024. [Link]

  • Anirudh Atmakuru, Alen Shahini, Subrata Chakraborty, Silvia Seoni, Massimo Salvi, Abdul Hafeez-Baig, Sadaf Rashid, Ru San Tan, Prabal Datta Barua, Filippo Molinari, U Rajendra Acharya, “Artificial Intelligence-based Suicide Prevention and Prediction: A Systematic Review (2019-2023)”, Information Fusion, 102673. [Link]

  • Anirudh Atmakuru, Subrata Chakraborty, Oliver Faust, Massimo Salvi, Prabal Datta Barua, Filippo Molinari, U.R. Acharya, Nusrat Homaira, “Deep learning in radiology for lung cancer diagnostics: A systematic review of classification, segmentation, and predictive modeling techniques”, Expert Systems with Applications, 124665. [Link]

  • Anirudh A, Lov Kumar, N.L. Bhanu Murthy, Aneesh Sharma, “Automatic Identification of Video Game Development Problems using Word Embedding and Ensemble Classifiers”, Innovations in Software Engineering Conference (ISEC), 2023. [Link]

  • Anirudh Atmakuru, Subrata Chakraborty, “Do Deep Learning Models Mimic Human Personality Traits? – An Empirical Study”, Australasian Conference on Information Systems (ACIS) 2022. [Link]

  • Anirudh A, Subrata Chakraborty, “Hidden and Face-like Object Detection using Deep Learning Techniques – An Empirical Study”, The International Conference on Digital Image Computing: Techniques and Applications (DICTA) - 2022. [Link]

  • Anirudh A, Aman Raj Singh, Anjali Goyal, Lov Kumar, N.L. Bhanu Murthy, “Prediction of Video Game Development Problems Based on Postmortems using Different Word Embedding Techniques”, Proceedings of ICON 2021: 18th International Conference on Natural Language Processing. [Link]

  • A Anirudh, “Mathematical modeling and the transmission dynamics in predicting COVID-19 - What next in combating the pandemic?”, Infectious Disease Modelling, 5, 366-374. [Link]

POSITIONS OF RESPONSIBILITY

Teaching Assistant - Elementary Real Analysis

August 2020 – November 2020
BITS Pilani
Instructor In-charge: Prof. Manish Kumar

  • Assisted the professor in preparing tutorial questions and invigilating exams for a class of 85 students.

Teaching Assistant - Probability and Statistics

January 2020 – April 2020
BITS Pilani
Instructor In-charge: Prof. PTV Praveen Kumar

  • Worked in a team of twelve professors and five teaching assistants and helped conduct various classroom activities, including course material preparation, technical assistance, and exam invigilation in a class of over 1,000 students.

SKILLS (GitHub - Link)

Programming Languages: Python, C, C++, Java, MATLAB, JavaScript, R

Developer Tools and Frameworks: Azure, Terraform, HTML, CSS, SQL, SPSS Statistics, Amazon Web Services (AWS), MongoDB, MySQL, Cassandra, PostgreSQL

Python Libraries: Pandas, NumPy, SciPy, Matplotlib, TensorFlow, PyTorch, scikit-learn, HuggingFace

Academic Achievements

  • Runner-up of the Computing and Electronics Research Summit paper presentation competition.
  • Awarded a scholarship of INR 10,000 for being among the top 5% of the performers at Ecom Express Pvt Ltd.

CO/EXTRA-CURRICULAR

  • Social Work: Was an active part of the NSS organization. Conducted free classes for underprivileged children and helped organize recreational events for old-age homes and orphanages.
  • Fest Participation: Organized various talks, student-faculty interactions, and events as a part of the Mathematics Association for the university’s technical fest - ATMOS.
  • Regional Association: Treasurer of the Tamil Sangam for the academic year 2019-2020.