NOUS logo

Innovations

Back to results page

HiveMind

Problems addressed

Centralised AI training violates strict data privacy regulations, incurs heavy network bandwidth overheads, and creates severe latency bottlenecks when streaming data to central servers

Unique Value Proposition

Modular backbone that abstracts away FL orchestration complexity (training rounds, aggregation, trust, scheduling), enabling multi-organisation collaborative AI training while raw data never leaves local infrastructure

Validated UC & Key Performance Benchmark

(Where was it tested, and what exact performance improvement or resource reduction did it achieve?)

Tested in: Integrated into ITML’s 3ACES analytics engine and validated on edge-to-cloud continuum inference testbeds.

Hard metrics:
-50% local memory and CPU compute overhead on edge devices
70% faster learning time
95% guaranteed delay compliance

Target Stakeholder Segments

(Who adopts/Hands-on Users)

AI/ML Engineers, Edge Software Developers, Telecom Infrastructure Leads, Enterprise Privacy Officers

Key Industry / Vertical Markets

(Where it is deployed)

Edge AI Infrastructure, Telecommunications, Distributed Data Analytics

License Model & IP and Commercial Strategy

(Open Source License)

Open Source License: Apache 2.0 built on Flower Framework

Commercial Streams:

Proprietary Strategy: Trade secret for Core IP

Availability

(Code Repository link)

Open Source License: Apache 2.0 built on Flower Framework

Commercial Streams:

Proprietary Strategy: Trade secret for Core IP

Estimated Time to Market

2–3 years post-project
Back to results page

Social Media

Linkedin Twitter Youtube

Other news items