Resume
Selected engineering work

Hybrid ML + RAG

Intelligent Patient Admission Triage Assistant

Deterministic ESI/MEWS scoring and ML risk prediction, separated from retrieval-grounded explanations.

PythonFastAPIPydanticLangGraphLightGBMXGBoostOllamaSentence-TransformersBGE embeddingsPostgreSQLpgvectorRedisCeleryDocker ComposeGitHub ActionsSyntheaFHIRRAGAS (planned)Langfuse (planned)Prometheus (planned)Grafana (planned)Loki (planned)
Ask about this project View repository
Choose your depth
Architecture, choices & results
01

The problem

Make the path from synthetic patient ingestion to deterministic scoring, risk prediction, retrieved evidence and explanation inspectable.

02

System architecture

Follow the flow from input to evaluated output. Select a stage to inspect its role.

Patient input — Patient input

FastAPI + PydanticLangGraphLightGBM / XGBoostOllamaPostgreSQL + pgvectorRedis + Celery
03

Engineering decisions

Engineering decision

Separate scoring from explanation

Problem
Keep deterministic ESI/MEWS scoring and ML risk modeling independent of generative explanations.
Approach
Separate scoring and prediction services from local Ollama explanation; consolidate structured data and embeddings in PostgreSQL + pgvector.
Trade-off / scope
Evaluation and observability remain designed/planned; no benchmark values establish performance or clinical effectiveness.
04

Evaluation, in context.

Designed evaluation and observability: designed. Synthetic patient/EHR-style Synthea / FHIR-compatible records; benchmark sample size not supplied.

05

Reliability & deployment

A documented safety and scope boundary.

In-progress research/portfolio system; synthetic Synthea / FHIR-compatible records, no hospital production deployment or clinical-validation result. RAGAS/Langfuse evaluation and Prometheus/Grafana/Loki observability are designed/planned. No completed benchmark values are supplied.

Documented services

FastAPI + Pydantic · LangGraph · LightGBM / XGBoost · Ollama · PostgreSQL + pgvector · Redis + Celery

Context limited to Patient Triage

Challenge this project.

Ask about the decisions, evaluation or boundaries behind the work.

Deterministic answers from documented portfolio evidence · project-scoped

A claim is only the beginning

Where did I use it?
What did I build?
How did I measure it?

Follow the projects, decisions and evaluation behind each answer. Every important claim has somewhere to go.

Direct · measured implementation Supporting · documented use Related · adjacent experience
An inspectable claimPG / 001
CAPABILITYRetrieval-augmented generation
IMPLEMENTATION

BM25+semantic retrieval

Trace the evaluation behind the number