Adrita Anika

Data Scientist with 5 years of experience in Artificial Intelligence and Machine Learning. Recipient of the Disney Data and Analytics Women Award 2022. Strong track record of delivering results and finding innovative solutions. Passionate about leveraging data to drive insights and make informed decisions. Excited to contribute my skills and expertise to a dynamic organization.

• Disney Data & Analytics Women (DDAW) Award, Disney Data Analytics Conference 2022, Orlando, Florida
• Dean's List Scholarship, Department of EEE, BUET Level 2, Level 3 and Level 4
• Techfest 2017, IIT Bombay India, (Asia's Largest Science and Technology Festival) 1st runner up, Digitalize Category.
• The Duke of Edinburgh's International Award 2018 The Silver Standard
• National Power and Energy Hackathon 2017, Bangladesh 1st runner up in the Smart Grid category.

Reviewer, Empirical Methods in Natural Language Processing (EMNLP) 2023

• Machine Learning
• Natural Language Processing
• Information Retrieval
• Human-Computer Interaction

Graduate Student Researcher, Texas A&M University, 09/2021-12/2023

• Multi-evidence Natural Language Inference (NLI) for Clinical Trial Data: Developed algorithms for determining entailment/contradiction given a hypothesis-premise (clinical trials) pair which requires reasoning. Explored several large language models (LLMs) for task-specific and domain-specific learning strategies. Additionally, explored evidence retrieval to support the decision of NLI through various ways of context representation including summarization, bag-of-words, BM25, Bi-LSTM, etc. Achieved 2-3% improvement in F1 score integrating context compared to non-contextualized case.

Student Researcher, Bangladesh University of Engineering and Technology, 01/2018-12/2020

• Search and Rescue with Drone-Embedded Sound Source Localization: Developed algorithms capable of localizing a sound source based on audio recordings made with an 8-channel microphone array embedded in an unmanned aerial vehicle (UAV) leveraging signal processing, machine learning and deep learning algorithms, published articles in international conferences and journals.
• Gaze for Responsive Interaction with 3D Avatar in Mixed Reality Environment: Developed a real-time system where 3D avatar can respond in a virtual environment based on the gaze-to-gaze interaction with the user leveraging machine learning and deep learning algorithms.

• DOANet: a deep dilated convolutional neural network approach for search and rescue with drone-embedded sound source localization, EURASIP Journal on Audio, Speech, and Music Processing, 2020, Qayyum AB, Hassan KN, Anika A et al.
• Fault Detection and Classification of Power System Busbar using Artificial Neural Network, 2019 IEEE International Conference on Power, Electrical, and Electronics and Industrial Applications, A. Anika, M. Junaed-Al-Hossain, S. Hasibul Alam, Nahid-Al-Masood
• Autonomous Trash Collector Based on Object Detection Using Deep Neural Network, TENCON 2019 - 2019 IEEE Region 10 Conference, S. Hossain, B. Debnath, A. Anika, M. Junaed-Al-Hossain, S. Biswas, C. Shahnaz
• Direction of Arrival Estimation through Noise Supression: A Novel Approach using GSC Beamforming and Room Acoustic Simulation, 2019 IEEE International Conference on Signal Processing, Information, Communication Systems, A. B. A. Abdul Qayyum, A. Anika et al.
• Estimation of Blood Glucose from PPG Signal Using Convolutional Neural Network, 2019 IEEE International Conference on Biomedical Engineering, Computer and Information Technology for Health, S. Hossain, B. Debnath, A. Anika, M. Junaed-Al-Hossain, S. Biswas, S. K. Zaman Navid
• Automatic Handwritten Words on Touchscreen to Text File Converter, TENCON 2018 - 2018 IEEE Region 10 Conference, B. Debnath, A. Anika

• Matching Patients to Clinical Trials

The objective of this study is to address the clinical trials matching problem: given a free-text summary of a patient's health record, locate clinical trials that are appropriate for that patient. We have utilized demographic filtering, relevance boosting with named entity recognition, neural re-rankers with BERT embeddings, adhoc query generation with generative LLM to address the challenges.

• Analysis and Visualization of Motor Vehicle Collision Data of New York City

 The project aims at exploring the motor vehicle collision data of New York City that can give insights regarding the causes of such accidents which may lead to potential remedies for preventing road accidents. Specifically, the project targets to explore the statistical relationship of road accidents with respect to the time of accident occurrence, types of injuries of the victim, vehicle types, etc. using multiple datasets containing more than 1M rows of data which are publicly available from Open Data, City government of NYC., Dashboard, Poster Bringing Machine Learning into the Classroom, The project aims at developing an interactive learning tool to help learners understand algorithms by step-by-step visualization. It has two main parts. The Sketch Viz enables users to learn sketch graphs e.g. speed graphs, curvature graphs etc. for each data point. The system's ML Algo Viz let users enter data points interactively, choose algorithms and hyper-parameters, run and visualize the model's training iteratively. User study results demonstrated that users found the interface easy to use & understand, it can help them learn, they will use it in the future and recommend it to others.

• Estimation of Public Speaking Anxiety from Bio-Behavioral data

The goal is to better understand individuals' affective responses while performing public speaking tasks. Several deep learning models, feature selection, feature transformation algorithms have been used on the Verbio Dataset.