Call for Applications - Apple Scholars in AI/ML PhD Fellowship

June 30, 2026

Call for internal submissions: Apple Scholars in AI/ML PhD fellowship

The Apple Scholars in AI/ML PhD fellowship was created to recognize and support PhD students in Computer Science and related areas who are pursuing research in artificial intelligence and machine learning, with a unique focus on work that is related to Apple’s core values.

Learn More:

Meet the 2026 Apple Scholars in AI/ML

Machine Learning research at Apple

Harvard University may nominate a total of three nominees, limited to one nominee from these research areas:

On-Device and Efficient AI

On-Device and Efficient AI focuses on making capable AI systems run on resource-constrained hardware - mobile devices, wearables, and custom silicon - without relying on the cloud. Research spans model compression, hardware-aware training, and compiler optimization, through to efficient foundation models, vision-language models, and streaming architectures designed for tight memory and latency budgets.

Sub topics: Model Compression and Efficient Architectures, Hardware-Aware Training and Inference, TinyML and Energy-Efficient AI, On-Device Foundation and Language Models, Efficient Vision-Language and Streaming Models, Large-Context On-Device Processing, Compiler and Runtime Optimization for ML

AI Security, Agent Safety & Assurance

AI Security, Agent Safety & Assurance focuses on protecting AI systems and the infrastructure around them from adversarial threats, misuse, and unintended behavior – from training time through deployment. This includes research on adversarial attacks and defenses, prompt injection resistance, data poisoning, and the containment of autonomous agents operating with elevated privileges or in high-stakes environments.

Sub topics: Adversarial Attacks and Defenses, Prompt Injection and Jailbreak Resistance, Data Poisoning and Training-Time Attacks, Agentic Containment and Risk Management, Privacy Attacks and Model Extraction, AI-Generated Content Provenance and Watermarking, Secure Deployment, Red-Teaming, and Supply Chain Security

Evaluation, Reliability & Measurement for AI Systems

Evaluation, Reliability & Measurement focuses on the methodologies, benchmarks, and frameworks needed to assess AI systems rigorously, consistently, and at scale. This includes designing evaluations for generative, multimodal, and agentic systems, as well as approaches to calibration, uncertainty estimation, and robustness testing that hold across diverse and out-of-distribution conditions.

Sub topics: Benchmark Design and Methodology, Evaluation of Generative and Multimodal AI Outputs, Human Evaluation and Annotation Frameworks, Calibration, Uncertainty, and Robustness Evaluation, Agentic and Long-Horizon Task Evaluation, Safety and Alignment Verification, Reproducibility and Scientific Rigor in AI Research

Computer Use Agents & AI-Assisted Software Development

Computer Use Agents & AI-assisted Software Development focuses on building AI systems that can plan, reason, and act across extended workflows - operating software interfaces, writing and understanding code, and completing complex developer tasks autonomously or collaboratively. This area spans foundation models for interactive agents, reinforcement and imitation learning, multi-agent coordination, and AI-powered code generation, verification, and debugging at the repository level.

Sub topics: Foundation Models for Interactive and Computer-Use Agents, RL and Imitation Learning for Sequential Decision-Making, Planning, Reasoning, and Multi-Agent Coordination, Tool Use and Long-Horizon Agentic Workflows, UI Automation and App Generation, AI-Assisted Code Generation, Synthesis, and Completion, Repository-Level and NL2Code Reasoning, Software Verification, Testing, and Debugging with AI

Human Centered AI

Human Centered AI focuses on designing, developing, and evaluating AI systems in ways that center human needs, values, and experiences throughout the entire development process. This includes research on conversational and multimodal interaction, human-in- the-loop learning, personalization, usable AI tools, and the emerging challenges of designing appropriate interaction paradigms for autonomous agents.

Sub topics: Multimodal and Conversational Interaction, Social Signal Processing and Human Factors, Human-in-the-Loop ML and Human-Guided Model Refinement, Model Personalization and Recommendation Systems, Usable AI Tools and Products, Agentic Interaction Design, AI for Augmented Cognition

AI Ethics, Inclusion & Social Responsibility

AI Ethics, Inclusion & Social Responsibility focuses on ensuring AI systems are fair, safe, and beneficial for all people - particularly those from historically underserved communities – as AI becomes more capable and pervasive. This includes research on bias mitigation, interpretability, guardrail mechanisms, value-sensitive design, and the assessment of sociotechnical impacts across diverse populations and deployment contexts.

Sub topics: Fairness, Bias Mitigation, and Inclusion, Interpretable and Transparent AI, Guardrail Models and Safety Mechanisms, Human-AI Interaction Risks, Value-Sensitive and Ethical Design for AI, Sociotechnical Impact Assessment

AI for Accessibility

AI for Accessibility focuses on developing AI systems that expand how people with disabilities interact with technology and the world around them, from communication and perception to cognition and navigation. Research spans augmentative communication, sign language recognition, accessible speech technologies, adaptive interfaces, and AI approaches tailored to cognitive and neurodivergent needs.

Sub topics: Augmentative and Alternative Communication (AAC), Sign Language Recognition and Generation, Computer Vision and Multimodal Interaction for Accessibility, Accessible Speech Technologies (ASR and TTS), Adaptive and Personalized Interfaces, Automated Accessibility Testing and APIs, AI for Cognitive and Neurodivergent Accessibility

Data-Centric AI

Data-Centric AI focuses on systematically improving the quality, composition, and management of data as a first-class concern in building reliable and fair machine learning systems. Research spans dataset curation and annotation, synthetic data generation, active learning, data selection and mixture optimization, and the attribution and valuation of training data.

Sub topics: Data Curation, Annotation, and Quality, Active Learning and Data-Efficient Methods, Data Augmentation and Synthetic Data Generation, Data Selection, Pruning, and Mixture Optimization, Data Attribution and Valuation, Fairness-Aware Dataset Construction, Data for Multimodal and Large-Scale Models, ML and Data Systems Integration

Privacy Preserving Machine Learning

Privacy Preserving Machine Learning focuses on developing techniques that enable AI systems to learn from and reason about data without compromising individual privacy. Research spans federated learning, differential privacy, cryptographic approaches including homomorphic encryption and secure multiparty computation, and privacy-preserving inference designed for on-device deployment.

Sub topics: Federated Learning, Differential Privacy, Private Inference and On-Device Privacy Mechanisms, Homomorphic Encryption and Secure Multiparty Computation, Cryptographic Primitives for Private ML

AI for Science, Health Signals & Biological/Physical Modeling

AI for Science, Health Signals & Biological/Physical Modeling focuses on applying machine learning to health and physiological data - from wearable sensors and time-series signals to longitudinal monitoring - to understand human health and support better decisions. Research spans foundation models for health domains, physiology-informed and causal modeling, behavior and activity recognition, and trustworthy clinical decision support grounded in continuous, real-world data.

Sub topics: Health-Domain Multimodal Foundation Models, Representation Learning for Sensor and Time-Series Data, Physiology-Informed and Causal Modeling, Longitudinal and Continuous Health Monitoring, Computational Modeling of Behavior and Activity, Trustworthy Clinical Decision Support, ML and RL for Mobile Health and Wearables

ML Theory

ML Theory focuses on developing rigorous mathematical frameworks to explain why and how machine learning algorithms work, to characterize their limits, and to guide the design of better systems. Research spans generalization, optimization, scaling laws, emergent behavior, and the theoretical underpinnings of modern large language models - including in-context learning, chain-of-thought reasoning, and test-time compute.

Sub topics: Foundations of Generative Models, Generalization and Out-of-Distribution Theory, Optimization Theory and Convergence Analysis, Scaling Laws and Emergent Behavior, Theory of In-Context, Few-Shot, and Reasoning in LLMs, Theoretical Foundations of Test-Time Compute, Foundations of Trustworthy AI (uncertainty quantification, calibration, robustness)

ML Algorithms and Architectures

ML Algorithms and Architectures focuses on developing the core models, training methods, and architectural innovations that advance the capability and efficiency of machine learning systems at scale. Research spans foundation and large language models, generative architectures, representation learning, retrieval-augmented systems, reasoning methods, and the growing landscape of alternative sequence architectures including state space models and mixture-of-experts designs.

Sub topics: Foundation and Large Language Models, Diffusion and Generative Models, Reliability and Factuality in Generative AI, Unsupervised and Self-Supervised Representation Learning, State Space Models and Alternative Sequence Architectures, Mixture of Experts and Long-Context Modeling, Test-Time Compute Methods, Reasoning and Chain-of-Thought in Language Models, Retrieval-Augmented Generation and Knowledge Grounding, Practical Optimization and AutoML, Multilingual NLP and Machine Translation, Interpretability and Model Representation

Spatial, Embodied & World-Model AI

Spatial, Embodied & World-Model AI focuses on developing AI systems that perceive, model, and act within three-dimensional visual and spatial environments - from mixed reality and computational photography to robotic perception and generative visual media. Research spans multimodal foundation models, neural rendering, 3D scene understanding, video reasoning, embodied AI, and sim-to-real transfer, with an emphasis on the challenges unique to spatial and continuous-world settings.

Sub topics: Generative AI for Visual Content, Spatial Computing and Mixed Reality, Semantic Scene Understanding, Video Understanding and Temporal Reasoning, 3D Scene Understanding and Reconstruction, Neural Rendering and AI-Driven Graphics, Multimodal (Vision-Language-Audio) Foundation Models, Vision for Robotics and Embodied AI, Synthetic and Simulated Data for Visual Learning, Computational Photography, Continual and Lifelong Learning, Sim2Real Transfer, End-to-End Spoken Language Models and Native Audio Models

All materials for the internal selection process must be submitted by noon on July 21, 2026. An internal faculty committee will review interested candidates to select the final nominees. 

Eligibility

Nominee must be enrolled full time at the nominating university, and expect to be enrolled through the end of the 2028/2029 academic year.
Nominee should be entering their last 2-3 years of study in Fall 2026. G1s are not eligible.
Nominee must not hold another industry-sponsored full fellowship while they are an Apple Scholar in AI/ML (Fall 2027 to Summer 2029)

REQUIRED MATERIALS for the Internal Submission:

CV, including name, email address, PhD program, and a list of all publications.
A research statement covering past work and proposed direction for next two years (two-page maximum, including citations, in a legible font size) clearly stating the hypothesis and expected contributions to the research area. Please note at the top the research area for which you would like to be considered.
A confidential letter of support from your thesis advisor (one page maximum).

Interested applicants should send their materials as a single PDF.  The confidential letter of support should be sent separately.

Please note: if you are nominated, the full application -- due on September 8 -- requires additional, more extensive materials than the internal submission.

All materials for the internal selection process must be submitted to: gsasacademicprograms@fas.harvard.edu by noon on July 21, 2026.

Questions regarding the internal application process should be directed to gsas-fwc@fas.harvard.edu