π Berkeley | π PhD Candidate @ UC Berkeley | π» Agent Evaluation | βοΈ Experimental AMO Physics
I build reliable AI systems and run trapped-ion experiments in atomic, molecular and optical physics. This page is a short snapshot of projects and papers I've been lucky to work on across both tracks.
- π§ͺ Agent evaluation and skills-based benchmarking
- π§ Deterministic environments for testing long-horizon workflows
- π οΈ Applied AI systems that operate across existing tools
- βοΈ FrontierPhysics - FrontierPhysics is an open benchmark measuring whether AI agents can carry out authentic, specialist-level physics research.
- π BenchFlow - BenchFlow is a frontier environment lab. We build the environments AI agents learn in. We ship SkillsBench, ClawsBench, PostTrain, and the runtime.
- π SkillsBench - SkillsBench evaluates how well skills work and how effective agents are at using them.
- π² first-tree (repo) - Open-source agent orchestration for engineers. Put Claude Code, Codex, Cursor and your own agents on one backlog β parallel runs on your keys, review before merge, everything lands as a pull request.
- π§ DeepTutor - DeepTutorZotero is a research sources manager based on Zotero, with amazing AI capability powered by DeepTutor.
- π mews - Represent you to finish all the work, when you are sleeping.
- π¦ smolclaw - High resolution mock environments for testing and improving claw like agents
- π SBTI CLI - SBTI CLI - Test SBTI for your agents.
- π NeurIPS 2026 Β· SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks
- π COLM 2026 Β· ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspaces
- π arXiv Β· BenchShield: Formal Model-Backed Instrumentation for Reward Integrity in LLM-Agent Evaluation Infrastructure
From control stack to integrated photonics, ion-photon interfaces, and microtrap design.
- π‘ Integrated photonics for trapped-ion systems
- π Ion-photon interfaces and multiplexed networking
- π§© Novel ion trap architectures, including 3D-printed microtraps
- π» bem - triangulation, boundary element method (BEM), fast multipole method (FMM) code for python
- π¦Ύ artiq_photonics_integration - ARTIQ Control Framework (ACF) of Photonics Integration
- π Nature Β· 3D-printed micro ion trap technology for quantum information applications
- π Phys. Rev. Lett. Β· Test of Causal Nonlinear Quantum Mechanics by Ramsey Interferometry with a Trapped Ion
- π Phys. Rev. Applied Β· Temporally multiplexed ion-photon quantum interface via fast ion-chain transport
- π npj Nanophotonics Β· Individual trapped-ion addressing with adjoint-optimized multimode photonic circuits
- π arXiv Β· A broadband, individually addressing two- and three-dimensional photonic integrated circuit for trapped-ion qubit control




