Leith, Edinburgh, United Kingdom
Real-Time Intelligence, Right at the Edge.
FreunDo develops low-latency, low-power FPGA-accelerated edge AI platforms for intelligent sensing systems, with an initial focus on next-generation hearing aids.
What we build
Hardware-proven AI for compact devices.
Modern hearing aids need advanced AI while staying tiny, private, responsive, and power efficient. FreunDo moves speech enhancement inference onto FPGA hardware so OEMs can evaluate and integrate deterministic, on-device intelligence without relying on cloud processing.
Solutions
From prototype to licensed acceleration IP.
IP core and reference RTL
Verified FPGA AI acceleration blocks for chip, SoC, and smart device manufacturers targeting real-time audio and sensor workloads.
FPGA prototyping board and SDK
Evaluation systems for hearing aid and medical device companies to test latency, power, audio quality, and integration constraints.
Partner integration services
Paid PoC projects, non-recurring engineering, model adaptation, quantisation, and reference firmware support for partner products.
Technology
A streaming AI pipeline designed for strict latency.
Mic arrays
AFE / ADC
FPGA preprocessing
Quantised NN
Post-processing
Receiver
RFD speech enhancement model
FreunDo's lightweight time-frequency model is designed for hearing aid constraints, with 24.1K parameters and approximately 40.3 MMAC/s computation in the BP benchmark.
- INT8 and INT4 quantised inference paths.
- Fully pipelined feature extraction, inference, and post-processing.
- Parallel processing elements for deterministic streaming performance.
- Local processing for privacy-focused, GDPR-aligned device design.
Market
Focused first on hearing and assistive devices.
Primary market
B2B hearing aids, assistive medical devices, hearing aid OEMs, ODM suppliers, audio SoC providers, and non-medical hearing device manufacturers.
Why now
AI hearing aids are moving from premium differentiation to mainstream expectation, while compact devices still face strict latency, battery, and processor limits.
Expansion path
The same low-power real-time inference platform can later support smart sports training, wearable health monitoring, industrial sensing, and security systems.
Our Team
Machine learning, DSP, FPGA, and embedded systems expertise.
Zechun Deng
CEO, Co-Funder, School of Physics and Astronomy, University of Glasgow.
Xin Feng
CTO, Co-Funder, School of Informatics, University of Edinburgh.
Graham Kerr
Business & Technology Consultant, Professor in Practice at the University of Glasgow.
Yi Tian
Chief Financial Officer, Master of Finance, University of Maryland.
Junhao Song
Tech Expert, Fully funded AI PhD, Imperial College London.
Usman Anwar
Tech Expert, Research Fellow, School of Engineering, University of Edinburgh.
Junchi Yang
Hardware Developer, MSc student, School of Engineering, University of Edinburgh.
Wayne Wan
Hardware Developer, MSc student, School of Physics and Astronomy, University of Glasgow.
Contact us
Build edge AI into your next device.
FreunDo is seeking OEM, ODM, semiconductor, hearing technology, and smart sensing partners for proof-of-concept evaluation and integration discussions.