New York, New York, United States
Current: Post-training Research @ MSL (Meta Superintelligence Labs) --- Previously: * MLE @ Google Gemini. Focusing on tool use * MLE @ Google Labs. Focus on Conversational AI / GenAI for Enterprise. * Tech lead @ Google Pay. I led multiple initiatives focused on discoverability and utility for merchants across Google Pay and Google Maps. * Tech lead @ Google Station. I led building a cloud-based (GCP) serving infrastructure for managing public wifi sessions. --- Interested in AI/ML, AR, VR, Computer Vision, Large Scale systems, Data Infrastructure/Pipelining, Experimentation design, Beautiful UI. Personal Interests: 3D Modeling, Jazz, Music Theory, Synthesizers, Music Production, Soccer, Basketball, Mountain Biking, and Hiking.
Post-training Research @ Meta Superintelligence Labs
Gemini tool use
Labs @ Google | 2023Q2 - 2024Q2 • Applied AI for Generative AI solutions (enterprise), building across the entire stack to merge research with product: Python, C++, TypeScript, SQL • Researched and built data infrastructure for creating augmented datasets from raw data sources to train (finetune) LLMs and improve inference quality • Researched leveraging NLP and NLU to analyze and evaluate model generated outputs for aligning with human preferences • Fine-tuning models for specialization, focusing on factuality grounding • Building information retrieval architecture (RAG) for LLMs with embeddings, vector index, search re-ranking, and various chunking strategies • Built a chat web app powered by an orchestrator managing a network of agents. Closely worked with product for rapid iterations to deliver a strong product review to executive leadership ------- Google Pay Web | 2022Q2 - 2023Q2 • Led the consumer feature pillar for a web surface across the full stack: Java, TypeScript • Drove workstreams to maximize cross-functional velocity with product and design teams for defining launch milestones for the zero-to-one launch of GPay Web • Spearheaded investigation over multiple legacy systems for scoping solutions to user privacy management. Led discussions among product, design, legal, and content teams to drive launch • Managed cross-team partnerships for a cluster of projects to establish collaboration opportunities, survey feature requests, and align on future roadmap for integrations
Google Maps Offers | 2021Q2 - 2022Q2 • As a TL, owned GPay Offers on Google Maps Mobile and led multiple launches while working across the full mobile stack: C++, Java, Obj-C, SQL • Owned the e2e execution of the Offers map pin caption, resulting in 32M DAU impressions and 20% increase in offers CTR. Spearheaded cross-team collaboration across multiple teams and systems while driving alignment through experimentation and systematic data analysis • Built a scalable distributed data infrastructure with Apache Beam to efficiently ingest, process, and index Offers data for serving on the Google Maps discovery surface ------- GPay US App | 2020Q2 - 2021Q4 • Led building a platform for serving merchant-focused features on the Google Pay app across the mobile full stack (Android & iOS cross-platform): Dart (Flutter), Java, TypeScript, JavaScript • Owned zero-to-one launch and growth of a multi-functional chat interface between users and merchants for surfacing various artifacts such as transactions, offers, and shipment tracking. Served 30M+ unique merchants and 150M+ transactions during ownership. • Prototyped and pitched a platform integration opportunity to a partner team for obtaining buy-in, then worked closely with cross-functional leads to define roadmap and execute ------- Google Station | 2019Q1 - 2020Q2 • Led launch of various features for a cloud-based architecture managing wifi network sessions by executing on the e2e stack: Go, GCP, Java, Kubernetes, Python, TypeScript • Spearheaded the design, deployment, and upkeep of a Kubernetes-based (Go binary pods) network traffic connector layer on GCP
• Developed internal user facing tools (full-stack) and backend infrastructure for serving critical risk data for the firm: Python, SQL (Postgres), Javascript • Built performant data processing pipelines with asyncio, tornado, and concurrent to parallelize tasks through multi-processing/multi-threading, achieving speed optimization of up to 80% • Implemented optimal data compute via vectorization using numpy, numba, and pandas, speeding up computation tasks by up to 56x
• Developed data processing and reporting pipelines for providing business intelligence: Python, T-SQL (MS SQL), SSIS, SSRS, VBA, C#, JavaScript • Delivered an ETL pipeline to generate and analyze time series data for revenue forecasting • Designed and developed a data warehouse with efficient ETL for flattened invoice and revenue data to increase query performance by 75% on average