Building a contract extraction pipeline to extract contract monitoring rules, figures, and relevant metrics from large-scale, complex contract documents.

Implementing functionalities for field suggestions, with human review, audit logging, and re-extraction jobs to keep extracted contract data accurate and compliant over time.

Natural Language Processing, under Prof. Adam Meyers (Spring '25, Fall '25, Spring '26)

Foundations of Causal Inference in Data Science, under Prof. Darryl Reeves (Fall '25)

Architected and implemented a system to convert natural language questions into executable SQL queries with automated visualizations and graphs using Python and LLMs, enabling non-technical users to build dashboards independently.

Unlocked $36,000 in yearly cost savings through the development and enhancement of this in-house analytics platform.

Designed a synthetic data generator for large low-dimensional datasets, which captured the distribution of the original surveyed data while ensuring that the generated data was differentially private.

Developed a software solution to compute differentially private statistics and histograms from a dataset. The solution also implemented different DP-adjacent anonymization techniques to release an anonymized dataset. The overall solution was deployed to over 50 metropolitan bodies across the nation.

Developed a rule-based approach that improved the existing process, leading to 21% more phones being found. Designed and optimized a pipeline for the end-to-end deployment of the model, reducing running time by 88%, and also ensuring data consistency and model explainability.

Prevented losses of an average $2 million/month by effectively identifying suspicious activity.