Lab

Range, without category blur.

Complete static content comes first. Filters progressively narrow the library, while every item keeps its maturity, evidence, and limitation visible.

8 projects

01
Maintained2026

LLM Steering

Local activation-steering and representation-engineering workbench with API, UI, tests, and experiment manifests.

Why it matters
Makes model-behavior experiments inspectable and repeatable.
Limitation
Research behavior does not imply general model control or safety.
PyTorchTransformersFastAPIReact
02
Reference implementation2026

Hospital Readmission FHIR ML API

Synthetic FHIR-facing API with deterministic demo scoring, explanations, tests, and clear safety boundaries.

Why it matters
Shows API and interoperability contracts without pretending a demo score is a validated clinical model.
Limitation
Synthetic deterministic scorer; not clinically validated.
FastAPIFHIRPythonDocker
03
Reference implementation2025

HL7 AI Challenge Platform

Standards-oriented public architecture for HL7/FHIR event flow, prediction services, care orchestration, and SMART/CDS surfaces.

Why it matters
Demonstrates interoperability architecture as a separate public reference.
Limitation
Not the Let’s Talk Doc award project and not represented as its winning repository.
HL7 v2FHIR R4RabbitMQCDS Hooks
04
Research2025

Local Document AI

Privacy-preserving PDF parsing, retrieval, local inference, and a lightweight review interface.

Why it matters
Provides public code for techniques adjacent to on-premises professional work.
Limitation
Public experiment; not a replica of an employer system.
OllamaRAGStreamlitOCR
05
Research2024

FreshTrack AI Module

OCR and structured-extraction service patterns for turning document images into typed outputs.

Why it matters
Explores document-to-data contracts and parsing failure states.
Limitation
Dependency-heavy research module; no real receipts or payment data are published here.
FastAPIOCRTesseractLLM parsing
06
Historical2024

ChatWithWiki Azure ML

Retrieval workflow assembled with Azure ML Prompt Flow over public knowledge sources.

Why it matters
Shows an earlier managed-cloud RAG implementation path.
Limitation
Historical demonstration; not actively positioned as a flagship.
Azure MLPrompt FlowRAG
07
Research2025

AlphaQuant

Local-first multi-agent research intelligence and forecasting architecture.

Why it matters
Explores orchestration, verification, storage, and evaluation outside the primary healthcare lane.
Limitation
Engineering research only; not investment or trading advice.
AgentsLocal modelsBacktestingReact
08
Archived2024

NFL Forecasting Archive

Historical model comparison, calibration, and backtesting experiments.

Why it matters
Preserves evidence of forecasting and probability-quality work.
Limitation
Archived analysis; not betting advice and not a maintained recommendation system.
XGBoostEloCalibrationBacktests