Computational science / systems architecture
From molecular motion to machine collectives.
I study complex systems by building models, infrastructure, and experiments around them. The scale has changed—from quantum dynamics measured in femtoseconds to GPU fabrics and persistent agent populations—but the practice remains grounded in first principles and evidence.
01 / scope
One method, five scales of inquiry
- Quantum matter
Quantum dynamics, molecular simulation, and ultrafast vibrational spectroscopy—connecting femtosecond-scale coherence and energy transfer to observable infrared and Raman spectra.
- Networks and resources
Energy-aware cloud provisioning and CPU/GPU algorithms for content-centric networking. The question shifts from molecular energy flow to computational work and information flow.
- Fabric as a system
RDMA, PCIe/CXL, memory pooling, remote devices, multi-rail HPC, and GPU fabrics—treating communication and memory movement as part of the computer itself.
- Computational science at scale
Reproducible, heterogeneous infrastructure for DNA and mRNA sequencing workflows, with each stage placed on the appropriate x86, Arm, or GPU resource.
- Machine collectives
Memory-efficient model runtimes, durable 1,000-agent infrastructure, and deterministic control planes for parallel work that remains observable, bounded, and recoverable.
02 / field notes
Recent measurements and arguments
- Scaling 1,000 AI Agents on OCI Kubernetes Engine and File Storage
A storage-backed runtime pattern for 1,000 persistent agents on OKE.
- Agent Graph 0.2.0: Parallel Issues, One Deterministic Control Plane
Parallel issue execution with leases, isolation, observation, and bounded control.
- Agent Graph: A Deterministic Control Plane for Multi-Agent Engineering
A control-plane architecture for software that contains agents inside its workflow.
- PCIe over the Network: When a Remote GPU Looks Local
Remote PCIe virtualization that preserves native driver behavior across a network.
- Raising an Agent: From Execution to Self-Evolution
Four capability levels: orchestration, reliable execution, workflows, and method discovery.
Writing archive
- The Contextual Turn: How AI Is Changing Our Understanding of Language
Language, context, memory, and interaction in AI systems.
- Agentify Cloud: When the Agent Becomes the Cloud Runtime
Agent-native cloud runtime design with policy, route intents, contracts, and validation.
- What Is Tau Computing?
Compute, memory, and fabric design for future AI systems.
- Zettascale in Practice: MRC Benefit
Systems-level thinking for AI networking and infrastructure efficiency.
- MRC and the Future of AI Networking
Network architecture for AI-scale communication and resource composition.
- Run Nextflow with heterogeneous computing on OCI
Pipeline-driven heterogeneous computing across Arm, GPU, and x86 cloud resources.
- Cut Nextflow costs by 70% with OCI
Cost-aware workflow placement for heterogeneous pipelines.
- Zettascale in practice: Scaling beyond limits
Infrastructure notes on scaling behavior, benchmarks, and large AI workload limits.
- Zettascale OSU and NCCL benchmark for H100 AI workloads
Benchmarking RDMA and GPU cluster communication for H100 AI workloads.
- The Story of OpenClaw: Learning to Collaborate
Human-guided AI engineering and system building.
- Human in the Loop: Guiding AI to Break the Bus/Network Architecture Barrier
Guided AI exploration for bus, network, and architecture design.
- Human-in-the-Loop Engineering and Vibe Coding
Human direction, agent execution, and engineering review.
- AI Coding BucketFS: A Transactional FUSE Filesystem for Object Storage
A storage and filesystem experiment shaped by AI-assisted implementation.
- AI Networking: How Fractal Scales Beyond Limits
AI networking notes on scaling, limits, and architecture tradeoffs.
- Inside AI Infrastructure, Series II: Benchmarking
Benchmarking as a way to reason about infrastructure behavior under workload pressure.
- A New Architecture Shift: NVIDIA, Enfabrica, and Intel
Accelerator fabrics and the changing shape of AI infrastructure.
03 / experiments
Systems under construction
- Agent Graph
Deterministic multi-agent engineering for VS Code: issue graphs, MCP control, isolated Codex workers, and parallel execution through bounded workflows.
Project pageInstall VSIX - Agent Runner
Server-side multi-agent orchestration with mailbox queues, agenda tasks, groups, resources, and deterministic project state.
Project page - Agentify Cloud
Agent-native cloud runtime with FastAPI, FastMCP, AGENTS.md policy, route intents, JSON contracts, and validation.
Project page - Memory-efficient MoE inference
Treats sparse experts as a managed working set: CPU memory holds the expert bank while routed experts move into bounded GPU slots for fused execution.
Passive offloadActive cache - Nextflow Infrastructure as Code
Reproducible DNA and mRNA sequencing workflows that provision stage-specific x86, Arm, or GPU infrastructure and retain task artifacts for inspection.
ArchitectureCost study
Project archive
- vLLM MoE CPU Offload
GPU-native MoE offload using CPU host memory as the expert-weight bank for constrained GPU memory.
Repo - llama.cpp-MoE
Router-aware GPU expert slots for local MoE inference under constrained GPU memory.
Repo - GPU-native Scheduler for GPU Computing
Patent-submitted scheduling approach for GPU resource management.
- SuperKernel for SuperPod GPU Clusters
Patent-submitted Jupyter kernel architecture for large GPU fabric execution.
Repo - Distributed MCP Protocol for AI-native CDN Architecture
Distributed protocol design for AI-native content delivery and agent coordination.
Demo - PCIe-Net and RDMA over PCIe/CXL
TCP/IP-over-PCIe/CXL and RDMA-style data movement across high-speed interconnect fabrics.
Demo - CXL Switch SoC and Cluster-on-Board
Switch-chip and multi-CPU board-level architectures for scalable AI and HPC systems.
- Multi-rail HPC Computing System
Multi-rail HPC computing architecture for production rendering workloads.
04 / record
A public trail across disciplines
Research
Seven archival papers trace a path through physical chemistry, cloud resource optimization, and high-performance networking. The topics differ; the common thread is quantitative modeling tied to computation.
- Infrared and Raman line shapes for ice Ih. I. Dilute HOD in H2O and D2O
- Infrared and Raman line shapes for ice Ih. II. H2O and D2O
- Vibrational energy transfer and anisotropy decay in liquid water
- On Energy-Aware Aggregation of Dynamic Temporal Demand in Cloud Computing
- Fast Longest Prefix Name Lookup for Content-Centric Network Forwarding
- Longest Prefix Lookup in Named Data Networking: How Fast Can It Be?
- Optimal Resource Provisioning for Dynamic Temporal Workloads in Data Centers
Speaking
Public talks connect many-core HPC, edge systems, telecom cloud, and AI networking.
Invention
Patent work spans named-data networking, rendering and ETL systems, AI applications, PCIe/CXL/UCIe interconnects, distributed storage, and GPU execution. Recent public examples include CXL-backed distributed storage and expandable PCIe address space.
05 / links