Hi, I’m Iyashi.
Messy input in. Action out.
I build AI systems that turn messy input into something a machine can act on — agentic workflows, RAG pipelines, and anomaly detection on real sensor data.
Data Scientist at Digestive Disease Care, one of two founding members of its AI and data science function · MS in Data Science, RIT.
The stack behind the work.
- Python
- PyTorch
- LangChain
- Google Gemini
- Claude
- PostgreSQL
- Redis
- Supabase
- FastAPI
- React
Workflows I've built
Granted
patent.
Cardiac Risk Prediction
A hybrid fuzzy inference model for multi-class cardiac risk. Granted as an Indian patent.
NGAFID Anomaly Detection
Predictive maintenance on 64-sensor aviation telemetry. 92.5% accuracy on NAB.
shadow360
A platform connecting students with clinical shadowing opportunities, co-built as a website and app.
What I work with
LLM Systems
- LangChain
- RAG pipelines
- MCP tool integration
- Multi-agent orchestration
- Claude, GPT, Gemini APIs
- Pinecone, Redis
- Prompt engineering
Machine Learning
- PyTorch
- scikit-learn
- Time-series modeling
- Anomaly detection
- Neuroevolution
- Feature engineering
- Model evaluation
- Computer vision
- MediaPipe
Engineering
- Python
- FastAPI
- Node.js
- PostgreSQL
- Redis
- Supabase
- Pinecone
- Docker, Git
- SQL, T-SQL, BigQuery
- Azure (Data Factory, Blob Storage, SQL, Key Vault, Entra ID)
- Power BI, DAX
- Entity resolution (rapidfuzz)
Ask about my work
A grounded assistant that answers from my project write-ups. It won't guess — if something isn't here, it'll say so.
Ask about a project, the stack, or the background behind it. Answers come only from this portfolio — if it isn’t here, it’ll say so.