Open Postdoctoral position, faculty mentor Philip Fisher

Stanford Institute for Human-Centered Artificial Intelligence (HAI); Palo Alto (CA); On-site; USD 81,000 - 99,000; Full time

Job summary

Stanford Institute for Human-Centered AI invites a postdoctoral scholar to join an AI-enabled instructional coaching initiative that advances AI methods for observing and improving high-stakes human interactions in education and related fields.

The fellow will develop multimodal models, ensure privacy and data governance, and collaborate with Stanford HAI researchers while contributing to real-world impact in early childhood settings.

Qualifications

• PhD in a relevant field with strong research or applied ML experience.
• Expertise in multimodal learning, CV, or speech recognition and deployment experience.
• Experience with human-in-the-loop machine learning or related feedback signals.

Responsibilities

• Design methods to convert coach interactions with AI-generated suggestions into usable supervision for model adaptation.
• Lead user studies evaluating AI-supported instructional coaching in real-world settings.

Skills

PhD in ML / Deep Learning

Multimodal learning

Human-in-the-loop ML

Large-scale models

Education

PhD in Computer Science / ML

Tools

AWS

Job description

Group Or Departmental Website

https://earlychildhood.stanford.edu/(link is external)

Faculty Sponsor First Name

Philip

Faculty Sponsor Last Name

Fisher

Stanford Departments And Centers

Graduate School of Education

Postdoc Appointment Term

One year

Appointment Start Date

Fall 2026

Does this position pay above the required minimum?

Yes. The expected base pay range for this position is listed in Pay Range field. The pay offered to the selected candidate will be determined based on factors including (but not limited to) the qualifications of the selected candidate, budget availability, and internal equity.

Pay Range

$90,000

Background

The Stanford Center on Early Childhood and the Stanford Institute for Human-Centered AI seek a postdoctoral scholar to join an AI-enabled instructional coaching initiative, which advances a broader agenda of AI for high-stakes social interactions. This work is motivated by a challenge pervasive in early childhood and beyond: in education, healthcare, and social services, the human interactions that matter most are often the hardest to observe, measure, and improve at scale. These settings also often involve vulnerable and underrepresented populations, making it essential to develop AI methods that can learn from real expert practice while meeting stringent privacy, consent, and data-governance requirements. The SCEC has developed iFIND, an interactive AI-assisted video editing platform that uses transcript-based models to help instructional coaches identify salient classroom moments and generate targeted feedback for educators. The postdoc will have access to a benchmark dataset describing social interactions involving young children, developed in collaboration with Serena Yeung-Levy's MARVL lab, as well as a dataset of videos collected through ongoing SCEC programs. This data will support the postdoc's work developing and deploying multimodal models to improve the iFIND tool, extending its current text-based approach.

Primary Responsibilities

• Human-in-the-loop model adaptation: The fellow will lead the design of methods that convert coach interactions with AI-generated suggestions, such as selections, edits, and rejections, into usable supervision for model adaptation. A central part of this work will be determining which forms of transcript, audio, and video data can be incorporated into training and fine-tuning in ways that are consistent with consent language, privacy protections, and applicable regulatory requirements. The fellow will develop and compare approaches for learning from sparse, noisy feedback generated during real use, including methods for multimodal signal integration, temporal identification of salient interactional moments, and model adaptation from implicit human feedback. The fellow will also build infrastructure to monitor model performance, calibration, and variation across deployment contexts and populations.
• A user study assessing instructional coach efficiency and fidelity with and without AI support: The postdoc will lead the design and execution of a study evaluating the use of their modeling approach in real early childhood intervention settings. This work may include usability and time studies, comparative evaluation of the efficiency and fidelity of coaching feedback produced with versus without AI assistance, and interviews with coaches about their experience using the tool.

Mentorship and Collaboration

Dr. Philip Fisher will serve as the official faculty mentor for this fellowship. The fellow will also work closely with SCEC Senior Machine Learning Engineer Dr. Lauren Klein Dubin, who will provide day-to-day supervision of the fellow's technical work. The fellow will have opportunities to publish with SCEC faculty and staff and present their work to academic, practitioner, and policy audiences, building a portfolio that demonstrates both methodological innovation and real-world relevance. In addition, the candidate will have the opportunity to engage with the broader Stanford HAI research community.

In addition to salary, this position will provide the following funding:

• $2,000 research/conference allowance
• Relocation support will be provided up to the following amounts: $3,000 (international), $2,000 (cross-country), or $1,000 (west coast, outside Bay Area)

Required Qualifications

• A PhD and a strong record of research or applied work in deep learning, with demonstrated ability to design, train, and deploy large-scale models
• Expertise in at least one of computer vision, speech recognition, or multimodal learning, with experience in real-world technology deployment
• Experience with human-in-the-loop machine learning, or a strong technical foundation to develop it, such as prior work with human feedback signals, weak/noisy supervision, or interactive ML systems

Preferred Qualifications

• Experience deploying models to cloud environments (e.g. AWS)
• Experience with or strong interest in privacy-preserving ML or human subjects research
• Passionate about child health and development

Required Application Materials

• CV
• Cover letter