Armada
Multi-agent orchestration framework for coordinating AI coding teams through phased contracts with evidence gates, role dispatch, and content collections.
Research · Engineering · Open source
Selected research and software, from information retrieval and evaluation to open-source systems.
A few starting points
Multi-agent orchestration framework for coordinating AI coding teams through phased contracts with evidence gates, role dispatch, and content collections.
World Bank toolkit for AI-assisted analysis of development data.
Model Context Protocol server for the World Bank Data360 platform.
Additional work and earlier contributions
Contributor to GLiNER’s relation-extraction utilities, including prompt-aware offsets, source-span indices, and separate entity and relation thresholds.
Conversational interface for querying World Bank Data360 data.
Proof-carrying numbers: verifiable claims about data provenance and computation.
Built and deployed Phoenix, Growsari’s company-wide JupyterLab environment, using Zero to JupyterHub on AWS with Terraform-managed, Kubernetes-based infrastructure. The platform is scalable and self-healing, and remains in use; its implementation repository is private.
Contributor to the open-source Python library for backtesting and optimizing machine-learning trading strategies. Contributions included data preparation, sentiment strategies, and backtesting improvements (2020–21).