01Healthcare
BM25 6.789FootCheck
FootCheck. Bringing technology closer to foot health. A platform focused on early awareness, self-monitoring, and education. Product engineering. Health education. AI workflows.
LIVE / AI ENGINEERING LAB
Intelligence you can
actually interact with.
Explore a working machine-learning system built into this portfolio. Search by meaning, inspect the evidence, and test the retrieval pipeline—all in your browser.
THE RETRIEVAL WORKBENCHKEYWORD ENGINE READY
Turn on real, local AI.
Enable a quantized language embedding model to find relevant work by meaning, even when the words differ.
Queries stay in your browser. No API key needed.First use downloads a ~23 MB model plus its runtime from Hugging Face and jsDelivr.
01 / ASK THE PORTFOLIO
TRY A QUESTION
13 portfolio sourcesDeterministic keyword retrievalAwaiting a search
02 / INSPECT THE EVIDENCE
3 sources for “Foot health and patient education”
02Clinical websites
BM25 5.887Eskander Foot and Ankle
Eskander Foot and Ankle. A multi-page practice website with condition education, hospital information, and new-patient guidance. Frontend. Content architecture. Patient experience.
03Security
BM25 1.814Cybersecurity engineering
Moheb is a cybersecurity engineer. His approach covers application hardening, least privilege, authorization, identity, patient-data protection, trust boundaries, and authorized hardware security research.
These are source excerpts, not generated answers. Ranking scores indicate retrieval relevance, not factual confidence.
01 Your question02 Tokenize keywords03 Rank sources04 Inspect evidence
03 / MEASURE, DON’T GUESS
Put retrieval to the test.
A small, transparent set of five portfolio questions. Run it against the loaded neural model and inspect every result.
Enable on-device AI above to run the evaluation.
Real model inference
A quantized transformer runs in a Web Worker, keeping the interface responsive while the model computes.
Private by architecture
Search text stays on your device. The model runs locally after its download, with no query API or stored conversation.
Evidence you can inspect
Each match links to its source. Export the retrieved context and compare keyword, semantic, and hybrid ranking.
Built with Transformers.js and MiniLM. This lab demonstrates the retrieval layer used in retrieval-augmented AI systems; it does not generate medical advice.
Explore the full engineering profileGOOD WORK STARTS WITH A CONVERSATION