Projects

Spring 2024  ·  CS 5302: Generative AI for Natural Language and Speech Processing

Developed an end-to-end application that interprets, translates, and vocalizes spoken language in real time, tailored specifically for patient–doctor conversations. The system aims to overcome language barriers in healthcare by chaining four components: automatic speech recognition → LLM + RAG diagnosis → machine translation → text-to-speech synthesis.

• Speech Recognition: Transcribes patient speech into English for downstream processing
• LLM + RAG: Fetches accurate, contextually grounded responses from curated medical documents
• Machine Translation: Translates the diagnosis into the patient's target language
• Text-to-Speech: Vocalizes the translated output for accessible delivery

Fall 2023  ·  CS 6314: Dynamic Programming and Reinforcement Learning

Trained a game-playing agent across increasingly complex Tic-Tac-Toe variants, from standard 3×3 to 4×4 and full 3D 4×4×4 boards. The core challenge was handling the exponentially growing state space as board dimensionality increased.

• Implemented Value Iteration, Temporal Difference Learning, and Deep Q-Networks (DQN)
• Extended from 2D 3×3 → 4×4 → 3D 4×4×4 with progressively refined state representations

Spring 2023  ·  CS 535: Machine Learning

Investigated the extraction of audio features (e.g., MFCCs, spectral centroid, chroma features) from raw audio signals, followed by applying and evaluating classical machine learning classifiers on the extracted representations.

• Explored mathematical backgrounds of selected ML methods and evaluated classification performance
• Benchmarked multiple classifiers and documented insights in a research-style report

Spring 2023  ·  CS 432: Introduction to Data Mining

Analysed a drug consumption dataset to answer the question: What factors drive drug consumption patterns in Connecticut, USA? Applied state-of-the-art unsupervised and pattern mining algorithms to draw data-driven inferences.

• Applied DBSCAN for density-based clustering of consumption profiles
• Used Apriori and FP-Growth for frequent pattern and association rule mining

Fall 2023  ·  ECON 438: Econometrics II

Used a German healthcare panel dataset to investigate: What factors determine the number of recent doctor or hospital visits? Addressed this using two complementary approaches, with careful attention to assumption verification.

• Tobit models on a single cross-section, with normality diagnostics
• Fixed and random effects panel models across multiple periods, with serial autocorrelation tests

Fall 2022  ·  ECON 330: Econometrics I

Conducted regression analysis on a primary-source survey dataset to investigate whether gender has a statistically significant effect on academic performance. Applied OLS with careful diagnostics to test standard regression assumptions.

• Designed and administered a survey questionnaire for primary data collection
• Verified OLS assumptions (homoskedasticity, normality of residuals, absence of multicollinearity)

Fall 2020  ·  EE 100: Engineering Laboratory

Built a software system capable of detecting different arrhythmia types from raw ECG data — an early introduction to biomedical signal processing that later informed my interest in medical imaging and generative models.

Spring 2021  ·  PHY 100: Experimental Physics Lab I

Used real-time captured video frames and image analysis tools (Tracker and ImageJ) to estimate the velocity of the International Space Station (ISS) from ground-based observations — combining physics, optics, and image processing.