AGENTS · GEO · SYSTEMS
Multi-agent research, geospatial orchestration, and companion tooling are all sections of this journal. Tap into the thesis fragments, platform summaries, and library reflections to see how the work threads together.
Thesis drafts and notes on MAS behavior; every card points to the original source material.
The Distributed Artificial Intelligence outline revisits how a DPS-inspired controller and an autonomous MAS ensemble can stay aligned while minimizing resource waste.
The MAS paper shows how negotiation, defeasible reasoning, and a central designer can keep agents benevolent without crushing their local expertise.
Domain adaptation notes ground future inference in the dA-distance, discrepancy metrics, and Rademacher bounds so that deployments behave predictably when distributions drift.
I am the Lead Software Developer driving the EO Exploitation Development Platform and the operator ecosystem that keeps it running.
GeoAnalytics Canada is a web-native environment for processing, analyzing, and visualizing Earth Observation data. The platform, operator, and supporting crates choreograph secure deployments, data ingestion, and observability across hybrid clouds.
Recent work hardens the VM pod experience—scoped sudo for apt/dpkg, consistent identity propagation, and PVC mounts that mirror IDE expectations keep operator tooling predictable.
A rewrite branch is replacing the Go backend, React/TypeScript frontend, and Go operator with a cohesive stack so every layer shares the same observability, deployment, and safety contracts.
Rust and Python libraries that power imaging, gateways, agent orchestration, labeled arrays, and STAC access.
A pure Rust TIFF encoder/decoder that handles headers, IFDs, tags, compression (PackBits + uncompressed), and multi-page archives so downstream systems stay in Rust.
The eo-vfs gateway serves an S3-compatible surface backed by Azure Blob storage, synthesizes virtual metadata, caches aggressively, and pre-fetches EO chunks for Dask clusters and Pod IDEs.
The Rust Agentic Orchestrator stacks NATS JetStream, SQLite WAL, and Ratatui to manage coding agents, with Doctor- and watcher-based guardrails keeping tool execution contained in a sandbox workspace.
Provides labeled n-dimensional arrays atop ndarray, pairing human-readable dimension names and coordinate labels with type-safe, parallel-friendly operations.
An async-first STAC client with fluent search builders, optional pagination/resilience/auth features, and strongly typed STAC objects for catalog interaction.
An MCP server that normalizes STAC search, collection/item retrieval, aggregations, and data-size estimation while offering both text and structured JSON outputs.