Loading...
Case Study
An Offline AI Decision Register for the Operating Theatre, Zero Data Egress
Clinical teams get an auditable record generated in the room, with confidentiality guaranteed by architecture rather than by policy.
Industry
Edge & Offline AI
Services We offer
Edge AI
Offline LLM / RAG
Medical Software
MLOps
About the client
A global medical-technology company building for the operating theatre.
Problem Statement
Operating theatres needed AI assistance that could transcribe procedures and produce a decision summary, but clinical data cannot leave the room and connectivity cannot be assumed. A cloud assistant was a non-starter.
The Challenges
Real capability, on-box
Running transcription plus an LLM and a vector store on edge hardware meant right-sizing the models and the pipeline to the device.
Auditable by design
Every entry in the decision register is timestamped and traceable, so the record holds up under review.
Approach
We designed for the edge and for confidentiality from the first line:
- A plug-and-play transcription module for the theatre's video feeds
- LLM-based summarization running fully offline with Ollama
- A vector database for retrieval-augmented generation over local context
- Edge-centric deployment on Docker over a hardened Ubuntu Core base
The Solution
The assistant transcribes the feed, checks it against local context through RAG, and writes a running decision register, a timestamped, auditable summary of what happened and what was decided, all on the device.
Because the models run on-box, the system works with zero data egress and stays available even when the network does not.
The impact we delivered
1 bytes
Clinical data leaving the room
1%
On-device, offline inference
Read more case studies
A Video Pipeline that Strips PII On-Device Before Anything Streams
Transforming Proprietary Medical Data to FHIR at Millions of Records a Minute
Unified Device Management for Hospital Networks with Azure Arc
Get Detailed Case Study
An Offline AI Decision Register for the Operating Theatre, Zero Data Egress
An on-device assistant that transcribes operating-theatre video feeds and generates auditable decision summaries entirely offline, using an on-box LLM and retrieval, so no clinical data ever leaves the room.
Capabilities
Product Engineering
AI, Data & Intelligent Systems
UX & Product Design
Experience Engineering
Cybersecurity
Compliance Engineering
US · Muscat · Bangalore · Puttur, ISO 13485:2016 · ISO 27001:2022 · ISO 9001:2015
© Copyright 2026 - LogicHive