Apr 2023 – Aug 2025 · 2 yrs 4 mos · Gainesville, Florida, United States · On-site

• Conducting research on Virtual Analysis Networks and Explanations for Social Sensing Analysis (VANESSA) with a focus on improving team cohesion using linguistic features. This project aims on analysing and modeling team states, for Mars mission.
• Collaborated with ARLIS at the University of Maryland to develop Information Competition Simulation, leading the topic modeling component of the project.
• Architected and Implemented State of the Art Topic Modeling algorithm : Fidelity, employing Keyword Extraction and Embedding clustering and an open-source Large Language Models (LLAMA-2). Published at NAACL 2025
• Employed keyBERT and sentence transformers for enhanced keyword extraction via cosine similarity, improving topic coherence.

Jan 2022 – May 2022 · 4 mos · Pune, Maharashtra, India

Jul 2021 – Jan 2022 · 6 mos

• Mentored by Prof. Miller to develop an AI-powered proctoring system for online examinations. Integrated OpenCV, Dlib, YOLOv3, and Darknet for real-time face detection, head pose estimation, and suspicious activity tracking. Implemented Google Speech-to-Text for speech recognition and anomaly detection. Designed a robust pipeline to enhance exam integrity through AI-driven monitoring and automated alerts.

Mar 2021 – Apr 2021 · 1 mo · Kanpur, Uttar Pradesh, India

• Developed a multiclass classifier for hate speech detection using a dataset of approximately 125,000 samples. Utilized BERT and ELECTRA encoding architectures and implemented an ensemble of baseline models to enhance classification performance.

May 2020 – Jul 2020 · 2 mos · Rajasthan, India

• Worked on detecting emotions in hindi-english code mixed data.
• Collected and created the dataset of about 1500 code-mixed sentences.
• Trained and evaluated the dataset on baseline models such as Sub-word LSTM, Bidirectional LSTM, Naïve Bayes with character n-grams and achieved an f1 score of 0.76