I build backend systems, AI infrastructure, and automations I use myself. That includes storage engines, shared GPU queues, tools for managing coding agents, and pipelines that bring me the news and job opportunities I care about.
I like owning the whole path from an idea to something that runs. I test what happens when it breaks, measure the tradeoffs, and write about what worked, what failed, and what I learned.
Currently a Software Engineer Intern at DR Financial, after working on AI platform systems at GMI Cloud and ML systems at Qualcomm Institute.
CS @ UC San Diego, class of 2028.
- Strata: A C++ storage engine with crash tests that check whether acknowledged writes survive process failures. Try the demo.
- amac: One place to see my coding agents, automations, permission prompts, and follow-up work, including from my phone.
- Winnow: Compresses LLM inputs and checks how much useful information survives, alongside token savings and latency.
- Taut: Reliable networking over UDP, with benchmarks showing where it beats TCP and where it loses.
- Franq: A Go message queue with disk-backed storage and crash tests.
- Tollgate: Lets a team share an LLM provider key with individual limits and revocation.
- GPU Broker: A training-job queue across cloud, spot, and lab GPUs, with checkpoints and interruption handling.
- decode-anatomy: A profiling study of where LLM generation time goes on an RTX A6000.
I use amac to keep track of the agents and automations behind my daily workflows.
- Morning Brief: Builds a personal newspaper from news and research sources, then delivers a PDF to Discord.
- Job Discovery: Finds early-career roles, filters them, removes duplicates, and emails a digest.
- TickVault: Records crypto order books and reports gaps and unverifiable data so readers know which windows to trust.
- noisy-rag-bench: Tests how document noise affects retrieval and whether answers can still be recovered.
The bugs that only exist after you deploy: what broke when I deployed Tollgate, and what I changed.
- Open source contributor to Awkward Array.
- Hackathon winner: projects include Vigil, Locus, Antibody, EduForge, and Pylon.
- DR Financial: Software Engineer Intern · Jul 2026 - Present
- Startup Incubator: AI Automations and GTM Engineer · Mar 2026 - Present
- UC San Diego: Technical Assistant · Sep 2025 - Present
- GMI Cloud: Software Engineer Intern, AI Platform · Jun 2026 - Sep 2026
- Qualcomm Institute, Engineers for Exploration: Software Engineer, ML Systems · Oct 2025 - Jun 2026
- Inovasity: Software Engineer Intern, Applied ML · Jan 2026 - Jun 2026
- Xyntopia (Taskyon): Software Engineer Intern, LLM Systems · Jun 2025 - Oct 2025
- Google x Project Basta: Code2Career Fellow · Feb 2026 - May 2026
- AWS Student Builder Club at UC San Diego: Technical Workshops and Projects Lead
- OpenAI Student Collective at UC San Diego: Lead
- Adobe: Student Ambassador
Languages: Python, C++20, Go, Rust, TypeScript, JavaScript, Java, C, SQL, Bash, HTML/CSS
Backend and automation: FastAPI, Spring Boot, Node.js, PostgreSQL, Redis, n8n
AI and data: PyTorch, CUDA, Triton, vLLM, NumPy, pandas, Arrow, Parquet
Web and infrastructure: React, Next.js, Docker, Kubernetes, AWS, Terraform, GitHub Actions
Also familiar with
Languages: Swift, Kotlin, Ruby, Scala, C#, R, MATLAB, Julia, Lua, PHP, Perl, Zig, Jac, VBA, PowerShell, Assembly
ML and inference: TensorFlow, JAX, scikit-learn, ONNX, LangGraph, Hugging Face Transformers, Ray, Optuna
Data tools: Spark, Kafka, Airflow, dbt, Snowflake, BigQuery, ClickHouse, DuckDB, Power BI
Platforms and tooling: GCP, Azure, Cloudflare, Modal, Supabase, Helm, Linux, CMake, gdb, Nsight, pytest, Playwright