A5060204591- Advanced Bandit Algorithms Research
- Mobile Health and mHealth Applications
- Reinforcement Learning in Robotics
- Smart Grid Energy Management
- Face and Expression Recognition
- Robot Manipulation and Learning
- Age of Information Optimization
- Face recognition and analysis
- FinTech, Crowdfunding, Digital Finance
- Artificial Intelligence in Healthcare and Education
- Congenital Heart Disease Studies
- Human Mobility and Location-Based Analysis
- Biometric Identification and Security
- Financial Literacy, Pension, Retirement Analysis
- Autonomous Vehicle Technology and Safety
- AI in Service Interactions
- Particle Detector Development and Performance
- Language and cultural evolution
- Artificial Intelligence in Law
- ICT in Developing Communities
- Guidance and Control Systems
- Speech and dialogue systems
- COVID-19 Digital Contact Tracing
- Transportation and Mobility Innovations
- Gamma-ray bursts and supernovae
Harvard University
2024
Google (United States)
2022-2023
Atal Bihari Vajpayee Indian Institute of Information Technology and Management
2019-2022
Indian Institute of Information Technology Allahabad
2020
The widespread availability of cell phones has enabled non-profits to deliver critical health information their beneficiaries in a timely manner. This paper describes our work assist that employ automated messaging programs preventive care (new and expecting mothers) during pregnancy after delivery. Unfortunately, key challenge such delivery is significant fraction drop out the program. Yet, often have limited health-worker resources (time) place crucial service calls for live interaction...
Mastercard, a global leader in financial services, develops and deploys machine learning models aimed at optimizing card usage preventing attrition through advanced predictive models. These use aggregated anonymized patterns, including cross-border transactions industry-specific spending, to tailor bank offerings maximize revenue opportunities. Mastercard has established an AI Governance program, based on its Data Tech Responsibility Principles, evaluate any built bought for efficacy,...
In many real-world applications of reinforcement learning (RL), deployed policies have varied impacts on different stakeholders, creating challenges in reaching consensus how to effectively aggregate their preferences. Generalized $p$-means form a widely used class social welfare functions for this purpose, with broad fair resource allocation, AI alignment, and decision-making. This includes well-known such as Egalitarian, Nash, Utilitarian welfare. However, selecting the appropriate...
In this paper, we present PRIORITY2REWARD a Large Language Model (LLM) based application which incorporates health worker preferences for resource allocation planning in public programs. LLMs are increasingly used to design reward functions on human Reinforcement Learning problems. We focus LLM-designed rewards Restless Multi-Armed Bandits, framework allocating limited resources among agents. the context of health, our approach empowers grassroots workers tailor automated decisions community...
This paper studies restless multi-armed bandit (RMAB) problems with unknown arm transition dynamics but known correlated features. The goal is to learn a model predict given features, where the Whittle index policy solves RMAB using predicted transitions. However, prior works often by maximizing predictive accuracy instead of final solution quality, causing mismatch between training and evaluation objectives. To address this shortcoming, we propose novel approach for decision-focused...
Underserved communities face critical health challenges due to lack of access timely and reliable information. Nongovernmental organizations are leveraging the widespread use cellphones combat these healthcare spread preventative awareness. The workers at reach out individually beneficiaries; however such programs still suffer from declining engagement. We have deployed SAHELI, a system efficiently utilize limited availability for improving maternal child in India. SAHELI uses Restless...
Robotics has proved to be an indispensable tool in many industrial as well social applications, such warehouse automation, manufacturing, disaster robotics, etc. In most of these scenarios, damage the agent while accomplishing mission-critical tasks can result failure. To enable robotic adaptation situations, needs adopt policies which are robust a diverse set damages and must do so with minimum computational complexity. We thus propose aware control architecture diagnoses prior gait...
The success of many healthcare programs depends on participants' adherence. We consider the problem scheduling interventions in low resource settings (e.g., placing timely support calls from health workers) to increase adherence and/or engagement. Past works have successfully developed several classes Restless Multi-armed Bandit (RMAB) based solutions for this problem. Nevertheless, all past RMAB approaches assume that behaviour follows Markov property. demonstrate significant deviations...
This paper studies restless multi-armed bandit (RMAB) problems with unknown arm transition dynamics but known correlated features. The goal is to learn a model predict given features, where the Whittle index policy solves RMAB using predicted transitions. However, prior works often by maximizing predictive accuracy instead of final solution quality, causing mismatch between training and evaluation objectives. To address this shortcoming, we propose novel approach for decision-focused...
Harnessing the wide-spread availability of cell phones, many nonprofits have launched mobile health (mHealth) programs to deliver information via voice or text beneficiaries in underserved communities, with maternal and infant being a key area such mHealth programs. Unfortunately, dwindling listenership is major challenge, requiring targeted interventions using limited resources. This paper focuses on Kilkari, world's largest program for child care -- over 3 million active subscribers at...
LLMs are increasingly used to design reward functions based on human preferences in Reinforcement Learning (RL). We focus LLM-designed rewards for Restless Multi-Armed Bandits, a framework allocating limited resources among agents. In applications such as public health, this approach empowers grassroots health workers tailor automated allocation decisions community needs. the presence of multiple agents, altering function can impact subpopulations very differently, leading complex tradeoffs...
Harnessing the wide-spread availability of cell phones, many nonprofits have launched mobile health (mHealth) programs to deliver information via voice or text beneficiaries in underserved communities, with maternal and infant being a key area such mHealth programs. Unfortunately, dwindling listenership is major challenge, requiring targeted interventions using limited resources. This paper focuses on Kilkari, world's largest program for child care - over 3 million active subscribers at time...
Abstract Harnessing the wide‐spread availability of cell phones, many nonprofits have launched mobile health (mHealth) programs to deliver information via voice or text beneficiaries in underserved communities, with maternal and infant being a key area such mHealth programs. Unfortunately, dwindling listenership is major challenge, requiring targeted interventions using limited resources. This paper focuses on Kilkari, world's largest program for child care – over 3 million active...
Abstract Underserved communities face critical health challenges due to lack of access timely and reliable information. Nongovernmental organizations are leveraging the widespread use cellphones combat these healthcare spread preventative awareness. The workers at reach out individually beneficiaries; however, such programs still suffer from declining engagement. We have deployed Saheli , a system efficiently utilize limited availability for improving maternal child in India. uses Restless...
Learning to communicate is considered an essential task develop a general AI. While recent literature in language evolution has studied emergent through discrete or continuous message symbols, there been little work the emergence of writing systems artificial agents. In this paper, we present referential game setup with two agents, where mode communication written system that emerges during play. We show agents can learn coordinate successfully using communication. Further, study how rules...
This paper addresses the challenges of face attribute detection specifically in Indian context. While there are numerous datasets unconstrained environments, none them captures emotions different facial orientations. Moreover, is an under-representation people ethnicity these since they have been scraped from popular search engines. As a result, performance state-of-the-art techniques can't be evaluated on faces. In this work, we introduce new dataset IIITM Face for scientific community to...
Restless multi-arm bandits (RMABs) is a popular decision-theoretic framework that has been used to model real-world sequential decision making problems in public health, wildlife conservation, communication systems, and beyond. Deployed RMAB systems typically operate two stages: the first predicts unknown parameters defining instance, second employs an optimization algorithm solve constructed instance. In this work we provide analyze results from first-of-its-kind deployment of system health...
The success of many healthcare programs depends on participants' adherence. We consider the problem scheduling interventions in low resource settings (e.g., placing timely support calls from health workers) to increase adherence and/or engagement. Past works have successfully developed several classes Restless Multi-armed Bandit (RMAB) based solutions for this problem. Nevertheless, all past RMAB approaches assume that behaviour follows Markov property. demonstrate significant deviations...
In 2020, maternal mortality in India was estimated to be as high 130 deaths per 100K live births, nearly twice the UN's target. To improve health outcomes, non-profit ARMMAN sends automated voice messages expecting and new mothers across India. However, 38% of stop listening these calls, missing critical preventative care information. engagement, employs workers intervene by making service but can only call a fraction enrolled mothers. Partnering with ARMMAN, we model problem allocating...
Mobile health programs are becoming an increasingly popular medium for dissemination of information among beneficiaries in less privileged communities. Kilkari is one the world's largest mobile which delivers time sensitive audio-messages to pregnant women and new mothers. We have been collaborating with ARMMAN, a non-profit India operates program, identify bottlenecks improve efficiency program. In particular, we provide initial analysis trajectories beneficiaries' interaction mHealth...
This paper addresses the challenges of face attribute detection specifically in Indian context. While there are numerous datasets unconstrained environments, none them captures emotions different orientations. Moreover, is an under-representation people ethnicity these since they have been scraped from popular search engines. As a result, performance state-of-the-art techniques can't be evaluated on faces. In this work, we introduce new dataset, IIITM Face, for scientific community to...
Robotics has proved to be an indispensable tool in many industrial as well social applications, such warehouse automation, manufacturing, disaster robotics, etc. In most of these scenarios, damage the agent while accomplishing mission-critical tasks can result failure. To enable robotic adaptation situations, needs adopt policies which are robust a diverse set damages and must do so with minimum computational complexity. We thus propose aware control architecture diagnoses prior gait...