ESRO Titlecard
Research Infrastructure  ·  HPC & Systems
Energy Systems Runtime Observatory
A lightweight, open source real time monitoring platform providing unified observability across local machines, Docker containers, and HPC clusters. Built to address the visibility gap that standard infrastructure tools leave open for scientific AI workflows running across heterogeneous compute environments.
Streamlit Docker SLURM psutil Scientific ML Open Source
Materials Copilot Titlecard
Research Infrastructure  ·  HPC & Systems
Building an Offline-First Literature Intelligence System for HPC
How I designed the CO2M pipeline inside Materials Copilot to run without internet access, survive preemption, and still give researchers something useful at the end. An artifact-materialization approach to scientific literature intelligence.
Offline-First SLURM Parquet Streamlit HPC Materials Science
More GPUs Slower AI Infographic
HPC & Systems  ·  Distributed Training
More GPUs, Slower AI: Why Scaling Hardware Can Break Your Training
Why adding GPUs often fails to accelerate training. Explores communication overhead, floating-point precision changes, barrier synchronization, and practical scaling efficiency tests that most teams never run.
Distributed Training GPU Scaling HPC Clusters Scientific ML Amdahl's Law
DER LLM Titlecard
LLM Deployment  ·  Energy Systems
From Docker to Multi-Cloud HPC: Securing LLM Workflows for Critical Infrastructure
How a dual-container architecture deploys large language models across AWS, GCP, and Azure for energy grid analytics, with automated security testing, ML-driven insights, and a path to HPC at national laboratories. Published at IEEE Cloud Summit 2025.
Containerized LLM DER Analytics FastAPI IEEE Published
AI Reproducibility Titlecard
Scientific ML  ·  HPC & Systems
Containerized ML Is Not the Same as Reproducible ML
Containers improve environment portability, but scientific AI results also depend on hardware, numerical behavior, runtime library choices, and domain validation. A practical framework for thinking about reproducibility in scientific machine learning across three levels of rigor.
AI Reproducibility Scientific ML HPC Containers
Assumption Mismatch Titlecard
Scientific ML  ·  HPC & Systems
Assumption Mismatch: What Scientific Algorithms Expect and What Systems Deliver
Scientific algorithms carry implicit contracts with the systems that run them. CI/CD pipelines, HPC schedulers, cloud orchestration, and quantum queues were not designed to honor those contracts. Explore the gap and how to bridge it.
Assumption Mismatch Scientific ML HPC Cloud
Quantum Serverless Titlecard
Quantum Computing  ·  Cloud
Quantum Serverless: Scalable Quantum Workflows
How serverless and containerization are enabling scalable, reproducible quantum computing workflows in the cloud. Explore hybrid ML and quantum use cases and modern orchestration patterns.
Quantum Serverless Cloud Containers