Grok + Semantic Search/Real Time Trends, Spring 2025

Distributed Training in Composer + Regression Testing/Cluster Diagnostics Infra

Runtime in the Mosaic Research Org

Contrary identifies and invests in the world’s top talent. The firm has helped create more than $10 billion in value through investments in companies including Ramp, Anduril, Zepto, and others.

Led 4 hours of office hours each week for Reasoning about Computation (Discrete Math)

Researching biases in Latin through contextualized, position-encoded embedding analysis with Professor Christiane Fellbaum.

Developed and productionized from scratch on the backend using Python:- Automated and AI-assisted Scope 1 and 2 emissions calculations, using MongoDB query heuristics and GPT vector indexing on company’s own data, allowing for 15 second long calculations for clients- A GPT categorizer that uses asynchronous calls, structured response retrieval, and manually edited contexts to categorize Scope 3 products in a company’s supply, able to categorize over 84,000 entries in 30 minutes- AI-assisted querying of scanned bill pdfs- AI Chatbot for FEMA risk analytics- AI-assisted report generation (WIP)- Excel-to-MongoDB converterIncreased emissions calculation and analytics efficiency by 80%.

Worked under Dr. Mickey Stone to characterize and optimize stable, pollution-reducing nanocatalysts. Calculated conversion and degradation rates with automated data processing from reactor measurements.

Designed and optimized eco-conscious, viable fungicide inhibiting the Cdc14 enzyme of Fusarium oxysporum and other fungal pathogens. Used computer modeling in MOE, python search algorithms, and in-vitro techniques.