Google
May 2026 – Aug 2026Software Engineer Intern · Python, TypeScript, Lit, Polymer, google3, Zipline, LLM Agents, MCPSan Francisco, CA
- Built agents on Jetski, Google's internal agent harness, that prototype UI directly against the internal web stack (Lit, Polymer, google3).
- Built a component knowledge graph: AST-parsed the TypeScript component library to extract props, types, and composition edges, indexed as a snapshot and served to the agents through an MCP server they query at generation time.
- Worked with and deployed across the Workspace, Gemini, and Google Search teams: built the pipeline for agents to deploy prototypes to a Zipline server behind a live preview link, migrated the core component library onto google3's new framework, and added MCP servers to the harness.
LinkedIn
May 2025 – Aug 2025Software Engineer Intern, Identity · Python, PyTorch, HadoopSunnyvale, CA
- Fine-tuned LinkedIn's internal foundation model with a contrastive objective over positive profile pairs, extending its embeddings to job history, titles, skills, and certifications.
- Ran Hadoop batch jobs generating embeddings across 500M member profiles, producing the vectors behind semantic persona search.
- Built the benchmarking harness (MRR, NDCG, AUC, serving latency) and measured a ~10% ranking-quality lift over the production baseline at unchanged latency, across the member recommendation and job matching pipelines.
USC Information Sciences Institute (ISI)
May 2024 – Sept 2024Research, Data-Center-Scale ML Infrastructure · Python, Rust, C++, GPU ClustersLos Angeles, CA
- Built "Git for models", a version-controlled registry for ML model checkpoints: content-addressed storage under a commit DAG, with branching and diff/merge across model variants.
- Designed the storage engine: chunk-level deduplication and delta encoding over serialized weights, cutting checkpoint storage 73% across the registry.
- Built multi-node sync on the shared GPU cluster that ships only the chunks a node is missing, cutting 12 TB/month of transfer; adopted by 8 research teams across the ISI lab.
Apple Inc.
May 2023 – Sept 2023Machine Learning & Software Engineering Intern, Apple Home · Python, C, TypeScript, PyTorch, React, FastAPICupertino, CA
- Built a hybrid Python/C ML library (ensemble models cross-validated against VAE latents to handle multicollinearity), reaching 94.3% accuracy on Apple Home device crash and proximity prediction.
- Designed backend services and telemetry pipelines behind internal APIs, cutting p99 serving latency from 78 ms to 22 ms, with model-metric dashboards and A/B comparison surfaces on top in React + TypeScript.
- Shipped a multi-threaded REST API, an analytics web UI, and the MCQI failure-metric system; presented root-cause findings to Apple Home leadership and the org VP.
NASA Marshall Space Flight Center (USLI)
Aug 2022 – May 2023Research Payload Engineer, "Eyes in the Cloud" · Python, C++, SDR, ReactHuntsville, AL
- Built payload comms for a sub-scale rocket: RF telecommand, a bespoke SDR receiver, a custom UDP server for real-time flight analytics, and onboard image processing at 4,600 ft apogee; integrated a BMP388 + IMU stack with a React ground-station UI, validated through launch and recovery.