×

HEROIC: A Geospatial Registry of Telescopes and Instruments

HEROIC is a web application that supports coordinated observations across observatories. It gathers data directly from observatory APIs into a geospatial registry of 35 telescopes and 51 instruments, spanning facilities such as LCO, Rubin/LSST, NOIRLab, LIGO, KAGRA, and Virgo.

My role: I am the backend developer and deployment lead of HEROIC, a full-stack Python/Django application with REST APIs, Docker, and CI/CD.

OPEN HEROIC

Related talk: Blast & HEROIC: Coordinating the Multi-Messenger Sky in the Age of AI (Gateways 2026).

×

Gathering and Visualizing Usage Metrics for SCiMMA's Hopskotch Service

Hopskotch is the publish-subscribe service of the Scalable Cyberinfrastructure to support Multi-Messenger Astrophysics (SCiMMA) project, built on Apache Kafka. Astronomers use it to exchange alerts about unexpected events in the sky.

To keep the service's operations and usage visible, a Python pipeline fetches data from Hopskotch and stores it in InfluxDB, and Grafana dashboards track system operations and user activity, at about 100,000 messages per day across 840 active topics. The dashboards support the project's operational reporting.

My role: I built the pipeline that ingests the data and the dashboards that visualize it. The ingestion code is open source. The production deployment on AWS lives in a private repository, so the dashboard below is a static snapshot of it.

Check the GitHub repository with the ingestion code here:

VIEW THE DASHBOARD (SNAPSHOT)

×

Data Sherpa: Retrieval-Augmented Chatbots for Science Domains

Data Sherpa is a retrieval-augmented generation (RAG) chatbot platform. Researchers ask questions in plain language and get answers grounded in a domain's own documents and records. It now serves multiple science domains:

• an assistant for the Dark Energy Survey (DES), over 1,367 indexed scientific papers;
• an assistant for the Morrow Plots, over long-term agricultural records from 1888 to 2021.

My role: I am the lead designer and developer of the platform. Among other things, I resolved its ingestion memory bottlenecks with buffered processing.

The Morrow Plots assistant is live and has its own card below.

×

Morrow Plots Assistant: Chatting with Agricultural Records from 1888 to 2021

An AI assistant, built on the Data Sherpa platform, that answers plain-language questions about the Morrow Plots and their long-term agricultural records, from 1888 to 2021. It uses retrieval-augmented generation (RAG), so its answers are grounded in the plots' own records.

My role: I am the lead designer and developer of Data Sherpa, the platform this assistant runs on.

How to try it: the assistant is live, and access is granted on request. Open it, ask for permission to register, and I will approve your account. You can also email me at baltierra@me.com.

OPEN THE MORROW PLOTS ASSISTANT

×

Terraform Templates for OpenStack: From the Dashboard to a Pipeline

Instructional templates for managing virtual machines on OpenStack with Terraform, in four use cases that each add one idea: a single VM; several VMs from one reusable module; floating IPs and data volumes that outlive VM rebuilds; and a GitHub Actions pipeline that plans on every pull request and applies on merge, with remote state. They grew out of real work managing a research group's VMs on NCSA's Radiant cloud, and they work on any OpenStack cloud.

My role: I wrote these templates, drawing on how I provision research infrastructure at the PRI with Terraform and GitHub Actions. The repository contains placeholders only: no production configuration, project names, or credentials.

Check the GitHub repository here:

×

Cairo Web: A Geospatial Archive for the African American Heritage Trail

A Django and geospatial web archive for the African American Heritage Trail project in Cairo, Illinois, built as a volunteer effort.

OPEN THE CAIRO MAP

×

Real-time Pipeline and Dashboard with World Bank Indicators

Developed as a Minimum Viable Product (MVP) for the Office of Evaluation and Oversight (OVE) at the Inter-American Development Bank (IDB): a functional prototype based on real use cases and datasets, designed to scale into a full product.

      This project delivers an interactive, web-based dashboard designed to give IDB's Office of Evaluation and Oversight (OVE) a clear, at-a-glance view of how key economic and social indicators have evolved across IDB's Borrowing Members Countries since 2004. Rather than burying stakeholders in raw data tables, it presents trends for each country in a simple line chart, alongside a quick-reference table of the most recent numbers and a breakdown of indicator values by year. Everything updates on-the-fly directly from the World Bank's API, so you're always looking at the freshest available data without delay.

      I have focused on nine high-impact indicators, all expressed as shares of GDP or absolute amounts where relevant:
      • Agriculture, forestry, and fishing as a percentage of GDP.
      • Central government debt as a share of GDP.
      • Exports and imports of goods and services (as a percentage of GDP).
      • GDP per capita growth (annual percentage change).
      • GDP per capita in PPP terms, in current US dollars.
      • Inflation in consumer prices (annual percentage change).
      • Net official development assistance received, in current US dollars.
      • Government revenue excluding grants as a percentage of GDP.

      With just a few clicks you can switch between countries, choose any of these indicators to see its twenty-year arc, or dive into a tabular snapshot for any given year. The chart's smooth curves and subtle gridlines make it easy to spot inflection points—say, when debt spikes or when export shares rebound—while the live data table provides immediate context without scrolling through multiple screens. For OVE's evaluation teams, this tool streamlines the initial phase of any policy assessment or program review.

Check the GitHub repository with more technical information here:

The live demo is temporarily offline while I move it to new infrastructure.

×

AppEval: An LLM-Based Chatbot with IDB Knowledge

Developed as a Minimum Viable Product (MVP) for the Office of Evaluation and Oversight (OVE) at the Inter-American Development Bank (IDB): a functional prototype based on real use cases and datasets, designed to scale into a full product.

      This project leverages AI to turn the Inter-American Development Bank's lengthy country-program evaluation reports into an intuitive, chat-based assistant. By embedding PDF documents into a vector database, and powering queries with an OpenAI language model (gpt-4.1), users can simply type natural-language questions—like "What years did the Paraguay evaluation cover?"—and instantly receive clear, concise answers drawn directly from official IDB documents.

      The result is a low-barrier chatbot, called "AppEval", that puts critical evaluation insights at everyone's fingertips, without requiring hours of manual searching or familiarity with technical databases. It answers questions using the Extended Country Program Evaluation reports for Barbados, Costa Rica, the Dominican Republic, and Paraguay.

      Integrating a tool like AppEval is particularly important for the IDB's OVE core functions because it transforms static evaluation reports into an interactive knowledge resource, streamlining how OVE and program teams access and interpret evidence: staff can bypass hours of PDF sifting to focus on findings and recommendations, accelerating internal reviews and collaborative discussions and ensuring insights more rapidly inform policy decisions and operational planning.

Check the GitHub repository with more technical information here:

The live demo is temporarily offline while I move it to new infrastructure.

×

Sayonara: What the Kyoto Protocol Left on the OECD Countries

An interactive data story on the first big environmental agreement of the new millennium. Using OECD data, it follows greenhouse-gas emissions in OECD member countries from 1990 to 2019, with a map, line charts, and treemaps, to ask what the Kyoto Protocol changed and which countries drive the totals.

My role: sole author, from the data analysis in Python to the story and the D3.js visualizations. Built at the University of Chicago, December 2022.

OPEN THE VISUALIZATION

×

People Shot and Killed by Police Officers During 2015

A single-page data story built with D3.js on the Washington Post's data about people shot and killed by police officers in the United States during 2015. Buttons switch between charts (The Overall Scenario, The Race Factor, The Race Issue, and The Ignored), each with its own short analysis.

My role: sole author, from cleaning the data in Python to choosing the charts and writing the D3.js code and the analysis. Built for the course Data Visualization for Policy Analysis at the University of Chicago, Autumn 2022.

OPEN THE VISUALIZATION