• Piloted ingestion of user impressions data for use in production machine learning models by tracking React component lifespans through efficient middleware.• Reduced impact of macro scale trends on lead to lease prediction as measured by the variance of positiveprediction rate by redesigning the model to a ranking system instead of a classification system.• Technologies leveraged: React, JavaScript, Python, FastAPI, pydantic, numpy, Nomad, Airflow, Git

• Reduced site latency on all pages significantly by building an event-based cache using Amazon’s EventBridge service and dynamic shape hashing for thrift objects.• Upgraded internal sales lead source to nearly double the number of properties visible to the sales team by upgrading a legacy external feed to the non-legacy version while avoiding all downtime in the sales funnel.• Reduced passive operating expenses from holding phone numbers by 500% via migrating the proxy phone system from Twilio to SignalWire, without dropping a single phone call to a landlord.• Technologies leveraged: Django, SQLAlchemy, Python, AWS, SQL, Git, Thrift

• Implemented automatic rent invoice creation to enable landlords on Zumper to collect rent in the same place they list their apartments and screen their tenants.• Reduced spam lead volume to nearly zero by deploying a transformer based spam classification model I designed.• Technologies leveraged: Tensorflow, TFX, Docker, FastAPI, Django, SQLAlchemy, Python, Git

• HackerRank (https://www.hackerrank.com/): Built pair programming into the Eclipse Che editor by implementing both conflict-free replicated data types (CRDT’s) and operational transformation (OT) based conflict resolution systems.• Zumper (https://zumper.com/): Designed and developed a model for automatic detection of spam messages by shrinking Google’s BERT model without sacrificing significant performance.• Polymorph (https://www.crunchbase.com/organization/adsnative): Optimized a model training framework to explore a variety of models to optimize price floors for profitability in real-time online ad auctions.• Technologies leveraged: JavaScript, TypeScript, TensorFlow, scikit-learn, numpy, pandas, Python, Git, Docker

ASE High School internship with IBM's Linux Technology Center. Full time for 9 weeks.