BEST SELLER 1
NON-INVASIVE DIABETES MANAGEMENT SOLUTION
The challenge was to build a non-invasive diabetes management system that measures blood glucose levels accurately using saliva.
Further, it should work in combination of saliva on a stick dipped in an enzyme pack that causes color change, and a mobile app that matches this color gradient to detect the glucose reading. The solution needed to work on all android phones (Lollipop, version 5.0 and above).
It was discovered that a Geometry-based approach, employed to scan the various parts of the “carton top” image, was complex and inaccurate.
To make the system simpler to implement & scale better across a large set of phone models, a Defined Open CV approach was adopted using Template matching.
The solution involved taking pictures such that it addresses variances due to different light conditions, shadows, different types of hardware. The captured image is processed using different formulas and arriving at a* value of the ‘target zone’ color schema and then the corresponding blood glucose value.
- Creation of a far more stable mobile app, cutting across phone models and operating system versions, courtesy the right ML implementation.
- Better CX through cutting down “recognition time” from 20 secs to 6 secs flat!
- Best-on-the-technology-landscape (Machine Learning approach) leveraged to tackle real world problems, with images.
- Just-in-Time design enabled client meet GTM deadline
- Rapid delivery cycle thus, helping client, with new feature releases, ahead of normal cycle
- iOS application and thus opening new vistas to a new set of users
BEST SELLER 2
PRIME – Patient Revenue Intelligent Maximizer Engine
Our client, part of one of Asia’s largest hospital chains, presented us with the challenge to stem revenue leakage due to lack of conversion of prescribed pathology and radiology tests to revenue. The Out-Patient to In-Patient conversion rate, too, was abysmal and was required to be significantly raised.
Our customized solution, PRIME, adopted an ML-based approach for prediction of probability of patient availing the service, by leveraging data like hospital billing, weather, distance from patient’s residence to hospital, load at hospital etc.
Importantly, PRIME was designed to intervene right at the point of outpatient billing, and was integrated with an “action layer” to send out the right communication.
- 150% increase in conversion, post implementation of PRIME
- Enabled hospital to intervene in real time, rather than post fact analysis
- Better understanding of customer segments ensured multi-fold increase in marketing effectiveness
- PRIME is live starting January 2018