Request for Proposal
Applied research award to solve complex Global Development Challenges using Artificial Intelligence and Machine Learning
About the Request for Proposals
AI and Machine Learning technologies have transformed the world. However, many underprivileged communities face serious barriers to good health, livelihood and comfort due to complex local and global development challenges. Today, AI and ML technologies have the potential to bridge this gap by providing solutions to many of these challenges and thereby bring the benefits of technology to those who need it the most. This Request For Proposals (RFP) aims to address this unmet need by inviting experts from across the world to apply their skills and knowhow to develop AI solutions for Social Good in partnership with the Wadhwani Institute for Artificial Intelligence.
We invite applicants to submit proposals that aim to solve complex challenges for the developing world by leveraging the power of AI and Machine Learning. We expect to award five grants of approximately USD 50,000 each, for one year. While we would prefer to fund proposed research in the areas of Wadhwani AI’s work in healthcare and agriculture, we will consider applications in other domains where AI can be employed to achieve scaled social impact for underserved communities.
Successful applicants will get a unique opportunity to implement their proposed research and work on-site with our team in India to develop solutions and deploy them by leveraging Wadhwani AI’s unique program capabilities. Applicants will be able to leverage datasets available at the Institute (subject to necessary permissions) or augment these with their data and other publicly available data. Furthermore, the opportunity provides a higher likelihood of social impact if combined and built upon the current projects that the Institute is involved in, although this is not a prerequisite for a successful proposal. Current domain areas of institutional interest are outlined below, and the projects are described in some detail (along with data availability) in the Appendix.
We are currently working in two areas where we expect AI-based technologies to have a high impact across the developing world: Maternal and Child Health and Infectious Diseases.
Maternal and Child Health
The first weeks of life have a disproportionate impact on a baby’s health. Timely interventions are crucial, especially for the 22 million low-birth-weight (LBW) babies born each year. Low birth weight contributes to 60-80% of neonatal deaths worldwide. If they survive, LBW babies are at risk of lifelong health problems. Most of these cascading poor outcomes can be controlled if the right care is provided early in the LBW newborn’s life, for which the baby must be weighed at birth. However, it is estimated that up to 50% of babies are not weighed at birth and many LBW babies are missed by the system. At our Institute, we are developing an AI solution that can accurately record anthropometric measurements such as weight and head circumference of babies using computer vision technology deployed through a standard smartphone.
Tuberculosis is the leading infectious cause of death globally – killing 1.5 million people in 2018. India harbours roughly 23% of the global burden of active TB patients. Gaps exist across the TB cascade of care, including lack of care-seeking among symptomatic individuals, missed screening and diagnosis, under-reporting of cases, lack of treatment initiation, and adherence to treatment. These gaps are particularly hard to overcome in the case of TB because of low awareness in rural areas, non-specific indolent early symptoms, poor quality of care in public and private sectors, and challenging treatment regimens. At our Institute, we are building AI solutions across the cascade of care for TB using a range of AI technologies and predictive models. Similar models may be applicable for other infectious diseases like HIV.
In addition to the above specific domain areas, Wadhwani AI also works on broadly demonstrating the value of contextual information to disease diagnosis and triaging. By contextual information we generally mean information that is not specific to the domain but provides value for the prediction task in the form of prior. Contextual information, for example, can refer to a patient’s medical history and comorbidities outside the condition being diagnosed. It can also include population and community level information such as socioeconomic and demographic factors, geographic information such as climate and season, and so on. Our aim is to build an AI and engineering platform that can broadly leverage context in healthcare in a scalable way.
Early Pest Detection
Our agriculture work is in the area of early detection of pests that attack cotton crops. In the year 2017-18, about 1,000 Indian cotton farmers, mostly in the State of Maharashtra, committed suicide because over 50% of cotton yield was wiped out by a Pink-Bollworm pest attack. Small-holder farmers, contributing to 75% of the country’s cotton production, struggle every year with uncertainty in yield and income. Many factors contribute to this uncertainty, but the inability to manage pests despite heavy pesticide usage is the most important one. At our Institute, we are building an AI solution using computer vision technology that enables cotton farmers to determine the type of pest infestation that may have taken place on their farm, estimate the density of pests by counting them on traps, and recommend countermeasures. The technology applies not just to pests on cotton farms but more broadly for pest detection.
Our solutions in each of the areas mentioned above are at various stages of continued development. We invite the larger AI research community to engage with us in building them to create scaled social good.
About Wadhwani AI
The Wadhwani Institute for Artificial Intelligence is an independent nonprofit research institute and global hub, developing and deploying AI solutions for large-scale social impact. It is, perhaps, the only organisation of its kind in the world that is entirely focused on AI for social good. By building AI-based solutions for vulnerable populations and
those at the bottom of the socioeconomic pyramid, it’s an unprecedented opportunity to bring cutting-edge technology to bear on very real social impact. Our approach is unique in how we work closely, on the ground, with
governments, international bodies, and nonprofit implementation partners, while at the same time pushing the boundaries of AI through world-class research.
At Wadhwani AI, we use the power of modern artificial intelligence to help solve some of the most challenging problems in the world. Our team consists of world-renowned scientists, innovators, entrepreneurs and domain experts in machine learning, healthcare, and agriculture.
Up to two members from each grant-winning team will get an opportunity to work on their proposed solution with the Wadhwani AI team at its India office. In addition to the grant monies, Wadhwani AI will arrange for or reimburse travel costs, as well as offer a stipend for the winners who will work out of the India office.
• Awards must comply with applicable Indian and international laws, regulations, and policies.
• Applicant must be eligible to receive grants as a Principal Investigator (PI) at an accredited academic institution that awards research degrees to PhD students.
• Applicant must be the PI on any resulting award.
All proposals shall be reviewed “Double-blind” and proposals shall be evaluated against the potential for impact.
Applications open 8 December 2019 and close at 23:59 (Indian Standard Time) on 7 March 2020. We expect to announce the winners of the grant during May 2020.
Submit Your RFP
Please use the form below to submit your application.
Additional information about the problems we are solving and the data available are described below.
Maternal and Child Health has been a predominant focus of health systems in developing countries such as India because the first 1,000 days of life – from the beginning of pregnancy to 2 years of age – are considered the most critical in a child’s development. For example, according to the WHO, stunting before the age of 2 years leads to poorer cognitive and educational outcomes in later childhood and adolescence. It also has significant educational and economic consequences for the individual, household and community. Within this period, the first month – the neonatal period – is the most critical. Low birth weight (LBW) contributes to 60%—80% of neonatal deaths worldwide. The neonatal mortality rate for LBW babies, i.e., babies with a birth weight under 2.5 kg, is 20 times greater than those born over 2.5 kg, and they have increased lifelong risk of diabetes, heart disease, reduced cognitive abilities, impaired immune function and stunted growth. Hence, identifying LBW babies and ensuring they get appropriate care has been a central pillar of health programs.
Every day, millions of frontline health workers worldwide visit homes of newborns for health assessments. However, vital measurements like weight are not accurately captured, due to issues in supply and maintenance of spring balances, cultural taboos that do not allow outsiders to touch newborns, low staff motivation, erroneous manual entries, and data tampering. Thus, too many of these LBW babies, who are easy to help, are missed.
We are building an AI-powered anthropometric tool that will enable frontline health workers to measure a baby’s weight, head circumference, and length in rural homes using only a short video taken on a generic smartphone, without network connectivity or additional hardware, and without requiring any views of the rear of the baby. We are currently conducting data collection to retrain the data, followed by field experiments.
Datasets available with the Institute
Smartphone-based Anthropometry: Support the development of an AI-powered anthropometry tool to capture additional health information/insights from a video of the baby. Available datasets include labelled baby pictures and videos, as well as a limited set of 3D scans of babies, also labelled. In addition, we have developed synthetic datasets of 3D baby models and software pipelines to construct additional synthetic data as necessary.
Tuberculosis (TB) is the leading infectious cause of death globally. India harbours roughly 23% of the global burden of active TB patients. The country has set a challenging target of reaching a total patient notification volume of 3.6 million by 2020, which is currently at 2.1 million cases and to eliminate TB across the country by 2025. Gaps exist across the TB cascade of care, including lack of care-seeking among symptomatic individuals, missed screening and diagnosis, under-reporting of cases, lack of treatment initiation, and adherence to treatment. These gaps are particularly hard to overcome in the case of TB because of low awareness in rural areas, non-specific indolent early symptoms, poor quality of care in public and private sectors, and challenging treatment regimens.
Many of these gaps maybe, to a small or large extent, addressable with the help of emerging AI technologies. Currently, AI has been limited to early use in radiology based detection, but less so in disease control programs. In our role as an AI research institute, we expect to build several AI solutions with the government data and other programmatic data from nonprofits, data collected by corporate entities, and our data collection efforts for the following challenges.
AI-driven adherence management: Proactive treatment adherence-related interventions (at the patient level, health worker level, and/or higher) integrated into workflows.
AI-powered active case finding: Directing case finding efforts to locations that are likely to yield a high number of TB cases.
AI-based vulnerability mapping: Map areas where populations are likely to be vulnerable to TB
Screening for TB: Develop a screening tool for TB by analysis of cough sounds and/or other data that can easily be collected at point-of-care.
Datasets available with the Institute
Central TB division notification and adherence data (expected), ground truth adherence data (expected), other survey and social media datasets.
Pest attack is a major threat to cotton farming. The consequences of yield loss are quite severe to the millions of small-holder farmers who get stuck in debt cycles and often take extreme measures. According to reports, a Pink Bollworm attack in 2017 destroyed over 50% of cotton in Maharashtra state alone, affecting over 80% of the 4.2 million hectares of cotton farms. This forced over 1,000 small-holder cotton farmers to commit suicide resulting in nation-wide grief.
The Government of India implements Integrated Pest Management (IPM), a scientific and holistic way of pest management, among small-holder farms through the State’s agriculture extension program. Under this program, the semi-skilled extension workforce (field staff) monitor pest infestation in a small number of demo farms in each block. Data collected from the demo farms is aggregated and analysed by the experts in agriculture universities who recommend pesticide interventions at the block level. Such a recommendation is often delayed and typically has a turn around time of a couple of days to over a week. The extension program faces several practical challenges: error-prone counting of pest density, delay in raising alerts and providing advisory and lack of adequate staff strength. IPM is both skill-oriented and knowledge-intensive, which makes it difficult to successfully involve uneducated/under-educated farmers and semi-skilled extension workforce in pest infestation monitoring. Similar programs are also implemented in other cotton-growing developing nations.
We have built an AI-based pest management solution that does pest identification and counting based on images of pest traps taken by farmers and extension workers and generates granular real-time alerts of pest infestation along with providing necessary advisory (recommendation) to the farmer. The solution thus helps with early detection and management of pest infestation.
For the solution we are targeting six major pests (that cause over 80% of crop damage for cotton crops):
- Pink Bollworm
- American Bollworm
- Sucking Pest
Datasets available with the Institute
We have a growing dataset of over 100,000 images of pests and pest traps collected from farms in India.