Wrote or rewrote astronomical pipeline tools for the Chandra Source Catalog.

• Optimized legacy algorithms for performance and reduced runtime by orders of magnitude.
• Worked in close collaboration with mission scientists to design and build new pipeline tools, turning science requirements into reality.
• Modernized 30-year-old legacy codebases to use modern C++ and Python idioms.

Concept-to-architecture effort for a modular, multi-archive science platform spanning CfA data centers and literature services.

• Prototyped a cloud-native platform that unifies access to cross-observatory Center for Astrophysics | Harvard & Smithsonian data within reproducible Jupyter environments.
• Established GitOps-driven infrastructure-as-code using Pulumi with dependency-injected components, enabling repeatable deployments and modular extension points.
• Evaluated and adapted Rubin Science Platform patterns to CfA needs, defining integration hooks for authentication, storage, TAP services, and collaborative compute workflows.

Unified the quality assurance toolchain so scientists can review intermediate catalog products rapidly.

• Built a reactive dashboard that responds to downloaded "QA"s, dispatches the appropriate web or desktop applications, and tracks file bookkeeping.
• Streamlined QA work both remotely and locally with a flexible configuration framework.
• Embraced the diversity of the QA tools and designed a smooth UX around it.
• Rewrote some of the QA tools for a better UX, better performance, or both.

Cloud-ready workflow runner that schedules, profiles, and debugs Chandra pipelines end-to-end.

• Grew a one-off script into a web service you can run on a laptop or a shared cluster without changing your workflow.
• Added queueing, telemetry, and timing hooks that surfaced bottlenecks and drove fixes with the science team.
• Gave developers a single place to run, debug, and compare pipelines, shrinking typical debug loops.

Modular software collection for building, orchestrating, and running toxicogenomics analysis pipelines end-to-end.

• Co-designed the overall workflow architecture described in the published paper, aligning reusable analysis components with real-world toxicogenomics pipelines.
• Led the design and coordination of a cloud user interface that lets scientists compose and launch Nextcast-based workflows without hand-writing orchestration code.

Turned a legacy ClearCase codebase into a first-class open-source package for X-ray astronomy analysis.

• Moved the project into Git, rebuilt the build chain, and started shipping wheels that install cleanly on Linux, macOS, and in containers.
• Layered in unit, integration, and regression tests with coverage checks so changes stop breaking long-trusted fitting routines.
• Sat with mission scientists to turn notebook prototypes into features that feel natural in their daily analysis flow.
• Worked closely with scientists to design and build new features, turning science requirements into reality.

Extensible desktop application for assembling, visualising, and modelling spectral energy distributions across multi-band archives.

• Integrated VO standards (VOTable, SED DM, SAMP) with Sherpa’s fitting engine so users could ingest, inspect, and model SEDs in one workflow.
• Built the plug-in framework/SDK that let partners ship custom services and desktop tools seamlessly.
• Harmonised heterogeneous photometry and spectra into a shared SED schema, easing data ingestion for VAO collaborators and future releases.

Virtual Observatory-compliant, web-based machine-learning platform for large astronomical surveys.

• Established a foundational science platform for data mining and exploration in Astronomy.
• Built browser-driven workflows that let researchers compose clustering, classification, and regression experiments on remote resources.
• Established an extensible framework for the addition of new models as well as the deployment of the platforms in different environments, from laptops to the GRID.
• Supervised computer science students for their BA theses.

Hybrid machine-learning method combining clustering with specialized regressors to estimate galaxy and quasar photometric redshifts.

• Delivered an efficient algorithm for photometric redshift estimation.
• Achieved improved robustness of results by minimizing bias and variance of the algorithm's output.
• Introduced deep learning concepts in the context of astronomical data analysis.