I've trained as a physicist and applied myself to quantum computing, a field where I frequently moved across hardware, software, and science and learned adjacent disciplines on the job.
That work led naturally into AI systems for research: agentic tools and harnesses, surrogate models, and cloud-backed workflows that make difficult technical work like building a quantum computer more tractable.
Currently based in the Los Angeles area.
Agent tools: Claude Code, Codex, Pi. Frameworks: PyTorch, JAX, OpenMDAO. Languages/platforms: Python, Julia, AWS.
EXPERIENCE
AMAZON WEB SERVICES
Sr. Applied Scientist
Jan 2023 – Present
[AI/ML] Built a memory plugin for agent harnesses that records session events as episodic memory, lessons from interacting with proprietary data and codebases, and distilled knowledge as readable Markdown. Used
qmdfor semantic matching and reranking. Reduced reprompting from every session to having the harness review an existing knowledge corpus before starting work.[AI/ML] Built a CI-based code review process that spawns a council of parallel sub-agent "persona" reviewers (pragmatist, security, end-user, etc.) with conflicting perspectives. Each persona reviews PRs independently against repository contribution standards and puts issues up for a vote by the rest of the council. Issues that exceed a specified threshold are sent to an auto-fix agent and the review process repeats until all issues are closed, reducing human review bottlenecks while maintaining engineering velocity.
[AI/ML] Built an autonomous agent loop using a tiered prioritization system with agents specialized toward literature, citation verification, exploratory research, knowledge curation, planning, execution, and introspection. Defined mission statements, artifact quality evaluations, and self-improvement mechanisms. As work is completed, the agent logs lessons, accumulates battle scars, and forms opinions on how to do future work. Used a Slack app as a desktop and mobile interface for high-level steering and overnight progress reporting.
[AI/ML] Built a self-prompting harness using a smaller model trained with DPO on a LoRA of Qwen 3.5 with training data sourced from prior session logs.
[AI/ML] Built an agent harness plugin that orchestrates high-performance compute (HPC) resources, manages compute jobs, and checks instance availability.
[AI/ML] Built reflection and work-distillation tooling that aggregated code harness logs, git activity, email, Slack, and desktop context to extract decisions, track commitments versus delivery, and generate agent guidance for multi-day prioritization and execution.
[AI/ML] Built a retrieval-augmented generation (RAG) assistant by scraping, downloading, and indexing over 100,000 technical documents to provide context for a large language model (LLM) answering questions against internal technical documentation. Used by technical staff to stay abreast of research across the org and by non-technical staff to make sense of jargon and research initiatives.
[AI/ML] Applied Gaussian process regression in PyTorch to create surrogate models of logical error rates given a set of physical error rates. Used surrogate model to understand system sensitivities to design parameters and as a building block towards global optimization of device parameters. Translated Torch models to JAX via ONNX for interop.
[AI/ML] Advocated for broader agentic AI use across the org by creating getting started guides, holding seminars, providing technical support, and consulting on specific applications of agents for multiple teams. Spun up EC2 instances to deploy frontier open-source models as an alternative when hosted model performance proved insufficient for some tasks.
[Quantum/Systems] Developed orchestration tooling to chain together numerical simulations. Enabled researchers to perform sensitivity analyses across solver functions sourced from multiple codebases (written either in Python or Julia) and run them in dedicated package environments on a scalable cloud cluster. This tool was used to compare the system-level architecture of two superconducting-based qubit platforms and the results were published here.
[Quantum/Systems] Built a library of subsystem models covering architecture definition, qubit physics and coherence, signal generation, cryogenic wiring and thermals, and quantum error correction. Integrated models using OpenMDAO to solve coupled disciplines and implicit functions to reproduce the experimental observation that qubit coherence can fluctuate over its frequency tuning range due to microwave reflections in the cryogenic signal chain.
[Quantum/Systems] Set up an automated qubit measurement testbed for engineers who did not have qubit measurement experience. Leveraged existing measurement automation infrastructure to create a "click and run" workflow to allow microwave hardware engineers to qualify their hardware with qubit coherence metrics.
[Quantum/Systems] Investigated the cause of signal delivery transients that exceeded the coherence time of qubits. Enumerated through all components in the signal chain and ran experimentally accounting for all contributions to pulse distortion. Used cluster-based FEM solver to run time and frequency domain simulations of chip and package. Proposed a simple circuit and verified the behavior in package-level simulations through geometry sweeps and trade studies on boundary conditions. Carried forth qubit experiments to measure and verify that the transient is eliminated with superconducting circuit boards.
[Quantum/Systems] Technical lead and advocate for device fabrication repeatability experiment. Oversaw the execution of data collection, organization, and analysis for a design of experiments. Created a physics-based model that halved the relative stdev. of Josephson junction across multiple wafer runs. This model predicted a source of variability caused by process equipment and not previously considered in design or operator training. Improved the algorithm for critical dimension estimation of junctions, increasing success rate from < 60% to > 99%, reducing the number of manual hours from about 2 to < 0.1 per wafer. This capability enabled repeatability monitoring of our most sensitive device dimension.
[Quantum/Systems] Forecasted thermal budget and constraints for a 1000 qubit system within existing commercial dilution refrigerators and identified technology development paths to further decrease thermal load and increase qubit count.
Quantum Infrastructure Manager
Aug 2021 – Dec 2022
Site lead for all quantum computing facilities. Hired, developed, and managed 12 direct reports, comprising scientists, engineers, and technicians. Promoted two team members and mentored an engineer through a role change. Coached team in writing and communication.
Launched an introductory series on quantum hardware to educate staff from adjacent fields, now used as onboarding content for multiple teams.
Spearheaded and completed three initiatives to ready research activities for productionizing quantum computers, bundled as one VP-watching goal: 1) a consolidated platform to ingest and display log data for dilution refrigerators, fabrication equipment, and facilities subsystems 2) a standardized runbook executed by trained technical staff for emergency and abnormal conditions 3) cloud access to quantum compute across AWS's network and security infrastructure.
Renovated an existing lab to ISO5 (class 100) cleanroom standards, which included an environmental enclosure (±0.1°C) for direct write lithography. Identified a cost effective water treatment solution to eliminate corrosion in the equipment chilled water system during building commissioning which saved almost seven figures in construction costs and months of delay.
Worked across AWS business units (real estate & facilities, data center engineering and operations, security, health and safety) to plan integration of quantum servers into AWS datacenters.
Applied Scientist
Feb 2020 – Aug 2021
Oversaw the design and renovation/ground-up construction of laboratory facilities totalling 40,000 SF and eight figures. For ground-up construction, transformed sketches and general program requirements from an empty parking lot to an operational facility with running cryogenic and fabrication equipment in 18 months. Delivered as a CEO-watching goal to open the AWS Center for Quantum Computing.
Served as a technical mentor for early science managers and trained staff on dilution fridge operations, cryogenic wire assembly, and device fabrication.
Managed vendors to configure, order, and coordinate the install of: the first three dilution fridges, an automated wafer prober, an automated wire bonder, a wafer dicing system, a flip chip bonder, and a plasma ashing system. Enabled core cryogenic, characterization, and packaging capabilities which later led to the development of a wafer-scale manufacturing and a 3D integration process.
Identified thermal and physical bottlenecks of cryogenic systems beyond 1000 physical qubits. Led conversations with commercial cryogenic vendors to discuss concepts to scale using emerging wiring and heat load requirements.
Filled in for technical product and program management roles. In a product role, served as the scientific editor for three (1, 2, 3) articles on the AWS Quantum Technologies Blog. In a program role, worked across functional managers and technical leads to define scope, perform resource analysis, and set a unified direction for 2021 research efforts.
EDUCATION
CALTECH
July 2015 - Jan 2020
Ph.D. in Applied Physics. Yariv-Blauvelt Fellowship. Advised by Oskar Painter. Built out a new part of the lab, designed and fabricated the first qubit, explored a high risk high reward fabrication method to make qubits more coherent when integrated with a transducer.
Thesis: Suspended Trace Air-Gap Resonators for Low Loss Superconducting Circuits
UC SANTA BARBARA
Aug 2011 - Jun 2015
B.S. in Physics and Applied Mathematics. Physics Academic Honors. Research Honors. Academic Excellence Award. Advised by John Martinis. Designed and made impedance matched infrared filters, cryogenic parts for Google Quantum AI's first custom dry fridge, and ran simulations for a tunable qubit coupler.
Thesis: Development of Hardware for Scaling Up Superconducting Qubits and Simulation of Quantum Chaos