1 of 23

AI biometrics from birth

Building inclusive, safe, and ethical tools to deliver healthcare

Simprints could become the first source of providing identity to hundreds of millions of children without identity now. It can allow these children access to services and protection from harm - something that they do not have today.

Sharad Sapra, Former Director of the Unicef Global Innovation Centre

❝

❞

2 of 23

Agenda

Problem

The global challenge
How can biometrics help?�

Solution

Hardwareless biometrics
R&D technical approach
R&D timelines�

Team & Sustainability

Biometric R&D and development sector expertise
Sustainability�

Impact

Short term
Long term

3 of 23

3

1 in 4 children born today lack any ID. This makes it incredibly hard to verify delivery of healthcare and aid.

For example, studies show 54% of children in Bangladesh don’t receive timely vaccinations despite official government estimates near 99%.

Gavi Chad audit 2019

“known differences between coverage rates as per administrative data compared to the survey data (approximately 50 percentage points)”
In 2015/16 alone, 700,000 pentavalent vaccines were unaccounted for
Link here

Gavi Rwanda audit 2018

Gaps of up to 42% at district level between vaccine stock levels and administrative data.
Every $1 invested in vaccines delivers a $16 saving in healthcare costs and lost wages and productivity due to illness. (Johns Hopkins Study 2016)
Link here

[Confidential] Document outlining research from Bangladesh MoH and other wastage data

In Bangladesh nationally there is at least a 10% gap between administrative data and WHO/Gavi assessments. In many districts it’s more. Assuming it has been more, not less, over time, then in a population of 164M this is over 16.4M people

4 of 23

Lack of reliable ID can lead to misidentification, duplication, and misreporting in healthcare delivery

4

Findings from an MNCH program serving 850k women through house-to-house CHW visits:�

Overall beneficiary duplication rate is 44.6%
Most of these duplicates come from 35% of frontline workers

Duplicates

Unique records

5 of 23

5

Biometrics can help solve this.

In a 3-year R&D project with Gavi, Simprints collected data from 15,000 children to build fingerprint tech for children 1-5 years old.

Today, this technology is being deployed to deliver >11M vaccinations in Ghana and Bangladesh.

Gavi Chad audit 2019

“known differences between coverage rates as per administrative data compared to the survey data (approximately 50 percentage points)”
In 2015/16 alone, 700,000 pentavalent vaccines were unaccounted for
Link here

Gavi Rwanda audit 2018

Gaps of up to 42% at district level between vaccine stock levels and administrative data.
Every $1 invested in vaccines delivers a $16 saving in healthcare costs and lost wages and productivity due to illness. (Johns Hopkins Study 2016)
Link here

[Confidential] Document outlining research from Bangladesh MoH and other wastage data

In Bangladesh nationally there is at least a 10% gap between administrative data and WHO/Gavi assessments. In many districts it’s more. Assuming it has been more, not less, over time, then in a population of 164M this is over 16.4M people

6 of 23

External evaluations show that biometrically-verified data drives real impact for mothers and children

6

Malawi: 56% increase in women linked to HIV care

Bangladesh: 39% increase in maternal health coverage

Ethiopia: ≥95% recorded patients received medicines, vs ≥68% in control districts

7 of 23

However, many interventions target children <12 months when they’re at their most vulnerable

7

Intervention	Age [months]
Intervention	0-2	2-6	6-9	9-12	12-16	16-24
Vaccines
Hepatitis B	HB1-HB3 \| DTPCV
Diphtheria, Tetanus, and Pertussis	DTPCV1	DTPCV2 - DTPCV3
Poliomyelitis	IPV1	IPV2 - IPV3
Haemophilus Influenzae Type B	DTPCV1	DTPCV2 - DTPCV3
Pneumococcal	PCV1	PCV2 - PCV3
Rotavirus	DTP1	DTPCV2 - DTPCV3
Meningococcal C and ACWY		MenC1			MenCA1	MMR/ Var1
Measles, Mumps, and Rubella					MMR1	MMRV2
Varicella					Var1
Human Papillomavirus
Malaria RTS,S		M1	M2-M3		M4
Nutrition / RUTF			RUTF	RUTF	RUTF	RUTF
Birth registration	Birth registration	Birth registration	Birth registration	Birth registration

8 of 23

This is a global challenge. The ability to verify vaccines, nutrition, & aid delivery is needed worldwide

8

In 2022 globally, 45 million children under five were wasted from malnutrition, stunting their growth and potential.

In 2022, there were 20.5 million children either un- or under-vaccinated. 1.5 million children died.

Of the billions of dollars the world spends fighting poverty, 29% of aid never reaches the people who need it.

In practice, the way we provide digital ID is through several biometric modalities such as fingerprint, face or palm recognition. We add this functionality to several data collection platforms which are currently used world-wide in delivering services, such as electronic medical records app or apps that track the delivery of money or goods in international development.

We have a hardware product (a fingerprint scanner) as well as a mobile app that has a simple user interfaces that guides the collection of the biometric images, runs a matching algorithm on the phone and issues a unique ID which is linked to the records. So for example if you are a healthcare worker looking to give the second dose of the covid-19 vaccine, with a swipe of a finger you could retrieve the record showing the first dose, without having to rely on names or other identifiers.

9 of 23

Biometrics <12 months is difficult due to the rapidly changing morphology of growing babies

9

Before recent advances in AI, biometric algorithms were unable to reliably identify children <12 months without expensive, custom-made hardware

In practice, the way we provide digital ID is through several biometric modalities such as fingerprint, face or palm recognition. We add this functionality to several data collection platforms which are currently used world-wide in delivering services, such as electronic medical records app or apps that track the delivery of money or goods in international development.

We have a hardware product (a fingerprint scanner) as well as a mobile app that has a simple user interfaces that guides the collection of the biometric images, runs a matching algorithm on the phone and issues a unique ID which is linked to the records. So for example if you are a healthcare worker looking to give the second dose of the covid-19 vaccine, with a swipe of a finger you could retrieve the record showing the first dose, without having to rely on names or other identifiers.

10 of 23

Any technology must also be ethical, privacy-first, & scalable in last-mile environments

10

Biometric

Template

Protected Template

We’ve partnered with the Red Cross to develop privacy-by-design approaches to digital ID

11 of 23

My starting point was seeing a 17 year old die of advanced HIV. So today we’re running the first trial in Africa of integrated HIV services. And because we don’t have to take patients’ names, their confidentiality is maintained which is very important when we’re talking about young people and sensitive services like HIV and reproductive health. ��Now we can advise the Ministry of Health which services are getting taking up for HIV care and what are the patterns of coverage. None of this would have been possible without Simprints.

Professor Rashida Ferrand,

London School of Hygiene and Tropical Medicine

Technology must also be designed for—and with—health workers, ministries, and decision-makers

❝

❞

Hear more user and patient stories our BBC Storyworks short film

12 of 23

Agenda

Problem

The global challenge
How can biometrics help?�

Solution

Hardwareless biometrics
R&D technical approach
R&D timelines�

Team & Sustainability

Biometric R&D and development sector expertise
Sustainability�

Impact

Short term
Long term

13 of 23

Our technical hypothesis is that multi-modal AI using convolutional neural networks (CNN) will work if we have enough training data

13

Costs and complexity of developing algorithms based on convolutional neural networks have decreased greatly

Recent evidence (e.g. Jain et al 2023) shows machine learning models can work with infants

The key limiting factor is training set size, something Simprints is uniquely positioned to do

In practice, the way we provide digital ID is through several biometric modalities such as fingerprint, face or palm recognition. We add this functionality to several data collection platforms which are currently used world-wide in delivering services, such as electronic medical records app or apps that track the delivery of money or goods in international development.

We have a hardware product (a fingerprint scanner) as well as a mobile app that has a simple user interfaces that guides the collection of the biometric images, runs a matching algorithm on the phone and issues a unique ID which is linked to the records. So for example if you are a healthcare worker looking to give the second dose of the covid-19 vaccine, with a swipe of a finger you could retrieve the record showing the first dose, without having to rely on names or other identifiers.

14 of 23

Our goal is to develop hardwareless biometrics that are accurate from the first day of a child’s life

14

R&D targets

Accuracy: >90%�
Size: ≤150 MBs�
Speed: ≤0.5s image processing�
Camera: ≤16 MP with autofocus�
Cost: Software only

In practice, the way we provide digital ID is through several biometric modalities such as fingerprint, face or palm recognition. We add this functionality to several data collection platforms which are currently used world-wide in delivering services, such as electronic medical records app or apps that track the delivery of money or goods in international development.

We have a hardware product (a fingerprint scanner) as well as a mobile app that has a simple user interfaces that guides the collection of the biometric images, runs a matching algorithm on the phone and issues a unique ID which is linked to the records. So for example if you are a healthcare worker looking to give the second dose of the covid-19 vaccine, with a swipe of a finger you could retrieve the record showing the first dose, without having to rely on names or other identifiers.

15 of 23

Our goal is to refine and deploy this algorithm in a 3 year project with existing sites in Ghana

15

	Complete	Y1	Y2	Y3
Algorithm refinement
Strategy and hiring	R&D Strategy Hiring engineers
Ethical data collection (10,000 infants)	Test methodology Establishing partnerships�Data collection software build	Collect data 1x per month to mimic vaccine schedule
Refine CNN algorithms		Cleaning up the data, image processing, extraction, CNN model, training, inference	Algorithm training and optimization	Continuous optimization
Field testing and piloting			Field testing	Continuous optimization
Deployment
Rollout to clinics	Continue existing rollout to >1 year old children in 586 clinics	Pilot v0 algorithm for <1 year old children in 10 clinics	Rollout v1 algorithm to 50 clinics	Rollout v2 algorithm to 586 clinics
Evaluate		Evaluate technical performance (accuracy)	Evaluate operational performance (user speed)	Evaluate impact (coverage)

In practice, the way we provide digital ID is through several biometric modalities such as fingerprint, face or palm recognition. We add this functionality to several data collection platforms which are currently used world-wide in delivering services, such as electronic medical records app or apps that track the delivery of money or goods in international development.

We have a hardware product (a fingerprint scanner) as well as a mobile app that has a simple user interfaces that guides the collection of the biometric images, runs a matching algorithm on the phone and issues a unique ID which is linked to the records. So for example if you are a healthcare worker looking to give the second dose of the covid-19 vaccine, with a swipe of a finger you could retrieve the record showing the first dose, without having to rely on names or other identifiers.

16 of 23

Agenda

Problem

The global challenge
How can biometrics help?�

Solution

Hardwareless biometrics
R&D technical approach
R&D timelines�

Team & Sustainability

Biometric R&D and development sector expertise
Sustainability�

Impact

Short term
Long term

17 of 23

We’ve got the combination of biometric R&D and development sector expertise to achieve this vision

17

Raj Kumar

Founder and Editor-in-Chief of Devex, industry’s leading media platform

Ted Dunstone

Founder of Biometric Institute
CEO of biometrics industry leader Biometix

Radhika Malpani

Former Google Senior Director
Creator of Google Image Search

Dr. Amir Hagos

Former Ethiopian Minister of Health
Global Financing Facility ($1.5B) country lead at World Bank

Board & Adviser highlights

Team highlights

Toby Norman

CEO

Harvard BA
Cambridge PhD
Forbes 30u30

Chris Royce

Director of Engineering

20+ years engineering
MBA U. Warrick

Tristram Norman

Chief Tech Officer

Forbes 30u30
MSc. Comp Sci.
Architected system >6.5M biometrics

Steve Taylor

Chief Delivery Officer

Led Syria response for World Vision
Oxfam, Save the Children AAH

Alexandra Grigore

Chief Product Officer

Cambridge PhD in Nanoscience
BioTech ‘Mover & Shaker’ winner

Eje Esangbedo

Director of Partnerships

19+ years sales
World Vision, Oxford, Pearson

Callum Woods

Senior Biometric Data Scientist

PhD Comp Sci.
Biometric ML specialist

Alfenur Kufa

East Africa Director

18+ years global health
Country Director, HPA Ethiopia

18 of 23

We have a unique advantage in access to frontline data combined with biometrics experience to gather AI training data

18

Expertise in “training” data collection, including ethical approvals

3x country offices with access to hundreds of thousands of patients

Experience collecting >6.5M biometric templates on the frontlines

19 of 23

Hardwareless algorithms will cost as little as $0.10 per patient to scale and sustain on the frontlines

Ghana

15,820 CHWs
$0.05 per person in a 33.4M population

Ethiopia

42,630 CHWs
$0.03 per person in a 123M population

Kenya

86,490 CHWs
$0.15 per person in a 54M population

Source: WHO’s Global Health Observatory Data Repository (African Region)

20 of 23

Agenda

Problem

The global challenge
How can biometrics help?�

Solution

Hardwareless biometrics
R&D technical approach
R&D timelines�

Team & Budget

Biometric R&D and development sector expertise
Budget�

Impact

Short term
Long term

21 of 23

Our near-term goal is to unlock scale to help deliver millions of doses of vaccines

2023

Ghana’s Minister of Health announces project to deliver malaria vaccines
586 clinics delivering 11M vaccine doses
Covering ~10% of the country

2027

Ghana’s president announces scale up for all vaccines in Ghana
9,909 clinics delivering 128M doses
Covering 100% of the country

22 of 23

Our long-term goal is to change the way the world fights poverty to protect children

22

In 2018 $4 billion in aid went to Yemen. This included twice enough food aid to feed every vulnerable citizen

Yet 85,000 children died of hunger and disease, amid reports of hundreds of millions of dollars of food aid stolen via fake beneficiary names

One example was Nasser Hafez, who died of hunger after his family could no longer afford to buy back stolen food aid openly sold on the local market. This should never happen again.

23 of 23

www.simprints.com

Our vision is a world where every person counts

toby@simprints.com