CTO at LlamaLab
Nandan Vernekar is CTO at LlamaLab. He has 25 years in software architecture and delivery, including solutions architecture roles in the UK and certifications in Azure, Scrum, and TOGAF. He writes about the systems behind same-day retrieval, OCR, and making the full medical record available to models without losing fidelity.
3 articles
Schema-first extraction was opinion encoded as infrastructure. The engineering that matters now is fidelity: guaranteeing the model sees the full record, structured so it can reason through it.
Records come back in 4 days on average, and 30 to 40% come back the same day. Same-day is a share, not a guarantee, and some states run faster.
How domain-trained neural networks and layout analysis achieve high accuracy on handwritten clinical notes, medical tables, and multi-generation faxes.