From Agents to Coordination: Reframing Distributed AI as a Systems Problem
What feels new about agentic AI is the tooling. What isn’t new is the problem.
ReadDeveloper Notes
Short posts about building and operating software. Mostly cloud systems, data pipelines, and geospatial work.
Written for future me, shared publicly. I hope something here saves you time.
Series: Agent Orchestration
Exactly-Once Is a Lie: Designing Replay-Safe Agent Workflows
Exactly-once execution is a myth. The only workable approach is at-least-once delivery with idempotent effects.
ReadAgent Systems Fail Quietly: Why Orchestration Matters More Than Intelligence
Most agent systems don’t fail because models are weak — they fail because coordination is underspecified, and the failures are silent.
ReadSeries: Librarian
Discovery as a Hidden Cost
Discovery isn't a one-time onboarding step; it's a recurring tax that compounds across agent-driven tasks.
ReadLatest
JIT SSH Certificates on k3d: Security Without the Friction
OIDC-backed SSH certificates, a Rust signing gateway, and Traefik multi-port routing for ephemeral workspaces.
Readeo-processor: Rust-Accelerated EO Processing from Python
A hybrid Rust + Python library for fast EO primitives (indices, masking, temporal stats) that plays nicely with NumPy, XArray, and Dask.
ReadProgrammatically Shutdown Uvicorn Server Running FastAPI Application
Running over localhost in a Kubernetes Job Pod - can be adapted for other environments
ReadAutomating the Creation of Georeferenced Sample Bounding Box Areas with Python
Georeferenced AOI grid-offset creation of bounding box sample areas.
ReadSaving DataFrames with Google Storage and Python
Saving a GeoPandas GeoDataFrame (also with Parquet and Feather examples) to a remote Cloud storage bucket and then read from the bucket - demonstrating Storage Client IO
ReadDask Futures: Do's & Don'ts
A few considerations with Dask Futures to improve computation efficiency and data management.
ReadSnow Mapping With MODIS
There are many reasons why snow mapping is beneficial. During Spring, we can use it to provide flood risk mitigation - especially for areas prone to flooding. During Winter, it cou
ReadISS Python Tracker
We will use the Cartopy library, available for Conda environments. This library allows us to project and transform appropriate mappings from coordinates and images.
ReadScheduling Tasks in a Distributed System
Curating the scheme of task allocation and its impact on the system
ReadStraggling Workers in Distributed Computing
The effects and approaches to consider when dealing with straggling workers
ReadAnomaly Detection Using PyTorch Autoencoder and MNIST
This article uses the PyTorch framework to develop an Autoencoder to detect corrupted (anomalous) MNIST data.
ReadApproaching Distributed Systems
The abstraction of distributed systems has become more prevalent as services like Amazon AWS , Google Cloud , and Microsoft Azure take the reigns away from developers and let them
ReadNo posts match that filter yet.