Academic research pipeline
Agent pipeline for the full research lifecycle: literature review, experimentation, statistical analysis, paper writing, and peer review. 8 skills, 58 agents.
Selected work
Research that ships.
Published papers and open code across the three pillars you just flew through.
Diagnosis
Models that read what clinicians see and hear.
-
Early Parkinson's from voice
Explainable voice analysis that detects early Parkinson's disease, published in Scientific Reports.
-
Chest pain triage
Multimodal models fusing clinical variables with free-text notes to flag critical chest pain in the emergency department.
-
LI-RADS categorization
Automated LI-RADS scoring for liver imaging reports.
Language
Clinical text, understood at the point of care.
-
Living evidence engine
Continuous ETL that discovers and ingests new open-access biomedical papers into a searchable knowledge base, with AI agents for retrieval and research gap analysis.
-
Live systematic reviews
AI-driven living reviews for the Brain-Heart Interconnectome, automating PICOS screening and evidence synthesis to cut research waste.
-
Auditable full-text screening
Screens full-text PDFs against review criteria with contrastive semantic highlighting and LLM judgment, every decision traceable.
-
Parkinson's self-care assistant
Bilingual assistant for people living with Parkinson's, built on a hard safety gate and answers where every sentence carries a validated citation.
Prediction
Risk models that see disease before it arrives.
-
Dementia risk from eConsults
Explainable NLP that surfaces early dementia signals in eConsult text.
-
Cancer label inconsistency
Explainable AI that catches inconsistent cancer classifications before they skew downstream models.
-
MASLD trial screening
Calibrated gradient boosting fused with LLM raters to triage thousands of liver disease trial papers, with auditable evidence spans.
-
Frailty from eConsults
NLP identification of frailty in provider-to-provider eConsult communication.
Research
Publications
The peer-reviewed work and preprints behind the systems above.
- Explainable artificial intelligence to diagnose early Parkinson's disease via voice analysis
- Benchmarking machine learning architectures for menstrual recovery prediction using physiologically informed synthetic wearable data
- From chaos to clarity: schema-constrained AI for auditable biomedical evidence extraction from full-text PDFs
- AI-driven live systematic reviews in the Brain-Heart Interconnectome: minimizing research waste and advancing evidence synthesis
- An auditable pipeline for fuzzy full-text screening in systematic reviews: integrating contrastive semantic highlighting and LLM judgment
- Inconsistency detection in cancer data classification using explainable AI
- AI co-scientist for knowledge synthesis in medical contexts: a proof of concept
- Graph-aware late chunking for retrieval-augmented generation in biomedical literature
- Explainable AI prediction of dementia in eConsults using natural language processing
Full lists on Google Scholar: Arya Rahgozar · Pouria Mortezaagha
Engineering
Built to ship.
Research only counts when it runs in production. The open-source infrastructure behind our systems.
Production AI ticketing
FastAPI system that creates, searches, and resolves tickets with RAG over historical and multimodal knowledge, exposed through MCP.
Natural-language API router
Drop-in FastAPI middleware that turns existing routes into a natural-language surface with LLM function calling. Zero new metadata.
Grounded document analysis
Async LLM document analysis with strict grounding and deterministic reporting, built for auditable outputs.
OCR ingestion pipeline
Asynchronous OCR extraction and annotation for PDFs at scale, feeding clean text into the evidence pipelines above.
E2E test generation
Auto-generates Playwright test suites from an app's source: page objects, fixtures, and tests for React, FastAPI, and more.
Agent mesh
Networked agent-to-agent messaging over HTTP and MCP: direct messages, broadcasts, threads, and permission relay between coding agents.
Alexa coding bridge
Voice control for a Claude Code session through Alexa: speak commands, get push notifications when the work is done.
Behind Larda
Prof. Arya Rahgozar
Co-founder · Professor of Data Science, University of Ottawa
Applies NLP and machine learning to medicine, predicting conditions like dementia, frailty, and osteoporosis from clinical data. Senior author across Larda's research, from living systematic reviews to the AI co-scientist.
Pouria Mortezaagha
Co-founder · AI researcher, University of Ottawa
Builds the models: retrieval-augmented generation over biomedical knowledge graphs, explainable AI, and clinical NLP, from auditable evidence screening to voice-based early Parkinson's detection.