Voice Biometric Platform
DeepInspire helped UK’s Biometric Authentication Company transform PoC into a large-scale enterprise solution.

Technologies:
AWS, C#, Node.js, React, NetCore, Jenkins, Bitbucket, Docker, MongoDB
About client
The client is a UK-based Biometric Authentication Company working within the area of biological artificial intelligence.
By unlocking the power of voice, using the latest machine learning techniques, the client reveals various market segments where the product can be used. One of the major segments is product usage through API by large enterprises (banks) in their business processes.

Location:
United Kingdom
Domain:
SaaS, Voice Biometrics


Challenges
The client, Biometric Authentication Company from the United Kingdom, needed a partner with deep custom software development expertise, who would help them transform PoC into a production-ready solution – a stable and maintainable Voice Biometrics Platform.
The main challenges the client faced were:
PoC version that works well for tens API requests but is required to handle hundreds of requests per minute.
API product only, without a client-facing component for analysing software performance and be able to tweak configurations.
Machine learning components built within the main business logic component (it leads to decreased performance and complicated development flow).
Machine learning components being sensitive to available memory and CPU resources, computationally heavy and becoming unstable on higher loads, no monitoring or ways to scale within PoC architecture.
Unpredictable PoC behaviour for any negative scenarios (only positive flow implemented).
The product had to be cloud-agnostic and be able to be deployed on-premises.
Mixed REST and RPC syntax in API (we had to standardise the API to enable maintainability and growth).
Looking for possible ways of computational optimisation and implementing them.
No deployment flow in place.
Team composition
& approach
We built three teams of DeepInspire experts, working hand in hand to ensure efficient delivery. In general, 16 specialists were involved in the project.
Team responsible for main business logic component
Experts:
-
Solution Architect
-
Node.JS Developers
-
DevOps Engineer
-
QA automation Engineers
-
Project Manager / Analyst
-
Project Manager Assistants
Team responsible for machine learning components
Experts:
-
Lead .Net Developer +
-
.Net Developers
Team responsible for analytical service application
Experts:
-
UX Architect
-
UI Designer
-
Front-end Developer
-
Back-end Developer
-
Daily team stand-ups
-
2 weeks sprints
-
Backlog grooming
-
Demos of developed functionality
-
Status calls
-
Meetups
-
Catch-up calls
Our solutions

01
Solution Architecture
Planning & Strategy
We designed and maintained the technological strategy/roadmap of the project. The work on the system was split into stages (e.g. moving Machine Learning parts to separate clusters, splitting into microservices, adding streaming etc.)
Architecture & Microservices
The system was divided into two clusters (business and computational). Machine Learning parts moved out of the business core into 3 separate microservices. Simple, easily maintainable and extremely robust. Analytical service was introduced in later stages as an independent project with its own back-end and front-end.
External API
Quality customer-facing APIs are extremely important since the primary user is another computer system consuming the API. We did complete refactoring of external API and created a beautiful intuitive and fully documented RESTFul API. Documentation was created, including the specifications themselves and developers documentation and guides.
Internal API
All services communicate via simple and fully documented internal APIs, including specifications and gRPC prototypes for each component (Business Component, 3 Machine Learning Services, Analytical Service).
02
Processes & Projects
3 Development Streams:
- Primary high-availability API;
- Machine learning microservices;
- A standalone Analytical Service with real-time
data analysis & reporting.
Development of Product & Project
Documentation including detailed user manuals
Release management and support operations for client testing environment

03
Development
Front-end development
Customer portal for managing and tweaking the system, analysis and real-time monitoring.
Back-end development
Including microservices, gRPC and REST APIs, webhooks, audio streaming & SDK for it, etc.

04
DevOps & Infrastructure
- The whole product infrastructure was set up in AWS;
- Since all DevOps processes and solutions were required to be cloud-agnostic, the whole solution is capable of being deployed to any cloud and on-premises as well;
- Design and development of product deployment, auto-scaling and instances hibernation based on current load.
- Design and development of project development flow and CI/CD using BitBucket, Jira and Jenkins.
05
QA / Testing
We designed and developed automated Quality Assurance using Postman (for creating a stable test automation environment, faster scaling and business expenses optimisation).
We helped transform PoC into a large-scale enterprise solution


Outcomes
-
Scalable, stable, easily maintainable and configurable Voice Biometric Platform, ready to be used by large enterprises.
-
Separate business and computational clusters with failover functionality.
-
Maintainable codebase.
-
Cloud-agnostic platform, able to be deployed on-premises.
-
RESTful API, fully documented, with guides.
-
Streaming, fully documented, with SDK and guides.
-
Standalone Analytical Service with real-time data analysis & reporting (to be able to see product performance in real-time).
-
Fully functional demo version of Analytical Service for ensuring client business development.