Biorouter AI Agent Marketplace

End-to-end research: scopes the question, reviews real literature, generates hypotheses, plans and verifies the analysis, then writes a cited draft. Delegates biomedical data to BioSkills and the SPOKE/CDW/OMOP extensions. It never fabricates a citation or result.

Stage-gated empirical research workflow: design register, data audit, identification, estimation, robustness, replication, and reporting.

Audits whether RCT, observational, IV, RDD, DiD, synthetic-control, matching, DML, or survival designs support causal language.

Checks research packages for clean-run reproducibility, relative paths, environment capture, seeds, logs, and table/figure mapping.

Verifies that manuscript claims, numbers, citations, figures, and tables are grounded in traceable evidence.

Confirms citations are real, relevant, current enough for the claim, and temporally valid.

Known-answer benchmark patterns for testing whether empirical-analysis agents avoid common causal and reproducibility failures.

Plans and verifies DOCX/PDF/PPTX/XLSX/LaTeX/scientific figure artifacts with source-to-exhibit mapping.

Reviews data/code availability, reporting checklists, protocols, ethics/privacy notes, and open-science readiness.

Plans scalable scientific data workflows with Polars, Dask, Zarr, TileDB, Arrow/Parquet, HDF5, and dataset versioning.

Designs and critiques statistical, Bayesian, survival, hierarchical, power, and uncertainty-reporting workflows.

Plans scientific ML with leakage checks, transformers, graph neural networks, active learning, SHAP, and reproducibility gates.

Handles geospatial analysis with CRS checks, spatial joins, rasters/vectors, maps, and spatial statistics.

Plans biosignal and neural-recording analysis: ECG/EDA/PPG, spike sorting, Neuropixels metadata, and peri-event QC.

Routes quantum and scientific simulation work across Qiskit, Cirq, PennyLane, QuTiP, and validation checks.

Guides computational materials workflows with structures, crystals, compositions, phase diagrams, and dataset provenance.

Plans DICOM/pathology/PACS workflows with tiling, metadata, QC, cohort provenance, and privacy-safe reporting.

Builds, cleans, verifies, and deduplicates Zotero/BibTeX/RIS/CSL citation libraries.

Plans Opentrons/cloud-lab workflows with plate maps, protocol simulation, sample metadata, and execution gates.

Plans applied econometrics in Python, R, or Stata: panel data, IV, DiD, RDD, clustering, and replication tables.

Runs systematic-review workflows: PICO, search strings, screening logs, PRISMA flow, extraction, risk of bias, and synthesis.

Anti-slop frontend design for landing pages, dashboards, lab tools, and paper companion sites. Reads the brief, picks a real design direction, and ships interfaces that don't look templated.

Plans discrete-event simulation, multi-objective optimization, sensitivity analysis, scheduling, and what-if modeling.

Routes astronomy, physics, units, coordinate systems, spectra, numerical experiments, and physical-science workflows.

Guides COBRApy/FBA metabolic modeling with GPR rules, media constraints, knockouts, objectives, and provenance.

Plans ADMET, virtual-screening, molecular-property, scaffold-split, leakage, and assay-provenance benchmark checks.

Reviews clinical decision support, treatment plans, and reports for evidence grounding, safety scope, and no-PHI handling.

Plans schematics, posters, slides, figure panels, infographics, visual abstracts, and source-to-visual traceability.

Structures research grants, specific aims, milestones, reviewer-risk responses, and funder-fit checks.

Reviews scientific software, device, and lab processes against quality, validation, traceability, and design-control gates.

Plans GPU-bound training/inference, batching, profiling, memory pressure, mixed precision, and reproducible compute.

Evaluates papers, scholars, venues, impact claims, reviewer fit, and publication positioning.

Plans qualitative coding, thematic analysis, codebooks, memoing, inter-rater checks, and audit trails.

Plans empirical finance/accounting workflows with event studies, abnormal returns, panel data, and table replication.

Plans Stata .do/.dta research workflows: audits, logs, tables, graph export, replication, and R&R updates.

Handles complex survey designs with strata, PSUs, replicate weights, domain analysis, calibration, and weighted inference.

Plans reliability, CFA/SEM, mediation/moderation, invariance, item analysis, and APA-style measurement reporting.

Plans dictionaries, topic models, embeddings, sentiment, supervised coding, LLM-label validation, and measurement-error checks.

Routes empirical econ/finance sources including WRDS, CRSP, Compustat, SEC/EDGAR, NBER, SSRN, and public datasets.

Plans journal targeting, referee reports, response letters, revise-and-resubmit execution, and appendix organization.

Routes game theory, equilibrium, DSGE/HANK, structural IO, calibration, simulation, and identification-proof workflows.

Plans CATE, causal forests, meta-learners, policy learning, conformal causal inference, fairness, and subgroup reporting.

Detects repeated workflows and drafts native Biorouter skills, hooks, checklist updates, or extension plans.

Runs evaluator-driven branch-and-prune exploration for improving research artifacts, prompts, analyses, and designs.

Finds and evaluates Hugging Face scientific models, datasets, Spaces, and papers with license and provenance checks.

Plans MD setup and trajectory analysis with OpenMM, GROMACS, AmberTools, MDAnalysis, MDTraj, and reproducibility checks.

Plans non-omics forecasting with classical models, ML, foundation models, backtesting, leakage checks, and uncertainty.

Audits BIDS neuroimaging datasets, derivatives, validation, metadata, and reproducible preprocessing readiness.

Plans clinical AI modeling workflows with EHR cohorts, validation, bias checks, calibration, safety, and no-PHI handling.

Plans CFD workflows with meshes, boundary conditions, turbulence models, solver validation, and convergence checks.

A pass/fail checklist for spotting AI writing patterns before generating prose, articles, or essays.

Applies ggplot2 best-practice style when writing R plotting code.

Applies tidyverse conventions and documentation standards when writing or reviewing R code.

Applies Python naming, typing, error handling, and project structure conventions when writing Python code.

End-to-end planning for the Ralph autonomous agent loop. Drafts a markdown PRD from a feature idea, then converts it to prd.json. Each story is sized for one hands-off iteration.

Engineering discipline skills: brainstorming, TDD, systematic debugging, parallel agents, plan writing, code review, git worktrees. Workflows tested on long-running engineering tasks.

SSH setup, SLURM job templates (CPU/GPU/H200), file transfer, module management, and common pitfalls for the UCSF CHPC cluster.