Open research / Code, models & evidence

Software that connects
questions to evidence.

Our contribution is an ecosystem: inference methods, physical models, reusable results, and the infrastructure that makes the work inspectable.

The group develops and contributes to collaborative scientific software. Some projects are inference engines; others connect tools, release data, teach methods, or test how agents can produce verifiable research. Each has a different role.

How the pieces fit

  1. Physical modelsSignals & kilonova surrogates
  2. InferenceRIFT & LISA-RIFT
  3. Workflow & reportingAdapters, tutorials & operating knowledge
  4. Reusable evidenceSamples, likelihoods & paper releases

Experimental foundation: campaign contracts + publication & verification
Make evaluations traceable, consolidation replayable, and claims independently checkable.

A map of contribution roles, not a claim that every project is already integrated into one pipeline.

Infer the source

From gravitational-wave strain to constraints on compact binaries.

Collaborative inference toolkit

RIFT

Estimate source parameters with an iterative likelihood-based inference framework.

Contribution
RIT and collaborators develop the inference methods and software; the official project is the starting point for current code.
Connection
Connects physical signal models to likelihood evaluations and posterior samples.

Research software

Official source Development lineage

Collaborator extension

LISA-RIFT

Extend RIFT inference to massive black-hole signals for LISA.

Contribution
Aasim Jan and collaborators adapt the RIFT framework to a different observing regime.
Connection
Builds on RIFT; demonstrates reuse of inference machinery across instruments.

Research extension

Source Associated paper

Model the physics. Reuse the evidence.

Surrogates make expensive models usable; public samples make published analyses inspectable.

Physical-model surrogate

Kilonova surrogates

Interpolate simulated kilonova emission for rapid modeling and inference.

Contribution
Marko Ristic contributes surrogate models and notebooks using simulations credited to Ryan Wollaeger and Oleg Korobkin.
Connection
Brings electromagnetic counterpart models into the multimessenger inference ecosystem.

Research models & notebooks

Source Associated paper

Reusable scientific outputs

RIFT public samples

Inspect posterior samples and marginalized-likelihood tables from published analyses.

Contribution
The group releases inference products with event-specific context and citations.
Connection
Provides downstream inputs for comparison, plotting, and population studies; cite each originating analysis.

Public data release

Data & citations

Collaborative data release

GW200105 samples

Explore the inference products accompanying a study of GW200105 eccentricity.

Contribution
A joint UT Austin/RIT release maintained by Aasim Jan; credit belongs to the listed research collaboration.
Connection
Links a specific scientific claim to reusable analysis products.

Paper-associated data

Data & paper

Connect the workflow

Turn individual programs into a research process people can run, inspect, and maintain.

Integration contribution

RIFT–PESummary ASIMOV adapter

Connect RIFT outputs to per-analysis and combined PESummary pages through ASIMOV.

Contribution
Our contribution is the adapter and its versioned RIFT asset interface. ASIMOV and PESummary are independently maintained upstream projects.
Connection
Separates reporting integration from the inference engine so each can evolve.

Integration package

Source & examples

Training contribution

RIFT tutorials

Learn inference workflows through worked RIFT examples.

Contribution
The group shares teaching and onboarding material alongside the scientific code.
Connection
Use these examples for orientation, and check the official RIFT project for current installation guidance.

Historical worked examples

Tutorials

Source-linked research practice

Operational knowledge

Carry practical lessons between projects, contributors, and agent sessions.

Contribution
RIFT integrator lore records sampler behavior and tradeoffs; SuperNu lore records operational recipes for the independently developed LANL transport code.
Connection
These are knowledge contributions, with source references and uncertainty labels, rather than new inference or transport engines.

Working documentation

RIFT integrator lore SuperNu lore

Make distributed work checkable

Experimental infrastructure for coordinating evaluations, preserving evidence, and reviewing claims.

Adaptive-workflow experiment

HyperPipe population demonstration

Explore adaptive inference and persistent evaluation caches on a synthetic population problem.

Contribution
A runnable demonstration of coordination and reuse across model evaluations.
Connection
Tests workflow ideas that may support larger inference campaigns; it is not a validated astrophysical model.

Synthetic demonstration

Source & demonstration

Shared interface experiment

Inference campaign contracts

Define evaluation records and replay-safe campaign assimilation.

Contribution
Experimental schemas, synthetic fixtures, and a reference reducer provide a small, inspectable coordination boundary.
Connection
A proposed connection between evaluators and campaign managers; science adapters and production adoption are not established.

Experimental · no stable API

Schemas & reducer

Publication & verification prototype

Trust and review research

Explore how agents offer evidence, check claims, rely on scoped findings, and amend errors.

Contribution
A public research dossier and agent-first prototype make the publication process itself available for inspection.
Connection
Supports the objective of autonomous publication with independent verification. Runnable exercises are distinct from an adopted or empirically validated protocol.

Research dossier & prototype

Research & prototype