• Designed and built a tool to monitor stability of Palantir’s third largest stack (5K+ live users). Surfaced 30+ opportunities for resource savings. Convinced resource management team to develop a similar tool for company-wide deployment.

• Built a typescript-based application for geospatial analysis of supply chain. Presented workflow to 30+ client stakeholders with huge positive response.

• Used Hugging Face’s open-source models to extract troubles and cause-corrections from 100+ unstructured PDF’s and word documents. Used the same models to recommend cause-corrections for new issues, reducing client’s trouble triage time.

• Built API endpoints to support credit card payments and interest disbursement.

• Owned a rent-estimation feature from product planning to back-end engineering in order to increase website traffic.

• Assisted company's migration from legacy infrastructure.

Conducted error analysis on output from Microsoft's Named Entity Recognition model on IU Health's doctors' notes, using Python Pandas and Scala in a Databricks environment.

Diagnosed issues in case-oversensitivity and lack of knowledge transfer between PII and health text model components.

Delivered novel insights on ways to collaborate with Microsoft to improve model performance to fit IU Health’s data.

Contributed to developing a Conditional Random Transformer model that searches chemical space to generate molecules that are similar to SARS-CoV-2 inhibitors, using Python RDKit, Tqdm, Pandas, and NumPy.

Analyzed Tanimoto similarity, Morgan fingerprints, pairwise similarity, and validity of molecules generated by CRT model using Python RDKit and Pandas.