Memories.ai Is Accepting Applications for Its 2026 Fall Research Fellowship

Memories.ai Research · July 2026

Memories.ai, the Silicon Valley AI company building the world's first Large Visual Memory Model, is opening applications for its 2026 Fall Research Fellowship. This is a 12-week paid research program for Master's students and early-career researchers working at the frontier of visual AI. Let's just say it's a great opportunity.

Fellows are embedded directly with the Memories.ai research team from September through November 2026. They work on unsolved problems in video understanding, multimodal memory, and real-time video intelligence — with full access to GPU infrastructure, proprietary datasets, and 1:1 mentorship from senior researchers. The program selects 5 to 8 fellows per cohort and provides a $1,000/month stipend. Visa sponsorship is available for international applicants.

Application inquiries: [email protected]

What Memories.ai Is Building

While major AI labs are investing in language model scale, Memories.ai is working on a different problem: giving AI systems persistent, structured memory for everything they see.

The company's core research direction is the Large Visual Memory Model (LVMM) — an AI architecture that can process, understand, and remember video at scale across long time horizons. Where current vision systems analyze a frame or a clip in isolation, an LVMM maintains retrievable memory of what it has seen over time — making it queryable, searchable, and useful as infrastructure for AI agents operating in the physical world.

Memories.ai's published research includes:

• SpatialMem — a spatial memory system that builds a queryable 3D index of indoor environments from standard egocentric RGB video, supporting natural language object and location retrieval without specialist sensors
• O-MARC — a compression distillation framework that makes omnimodal audio-visual inference 34.6% faster and 34.7% more memory-efficient while improving accuracy over full token inference
• OmniRetriever — the first model to unify audio, video, and text in a single retrieval space, surpassing Gemini Embedding 2 by up to 18.0 R@1 on audio retrieval benchmarks, zero-shot
• MARC — metric-aligned video representation for structured scene understanding

The fellowship is the entry point for researchers who want to contribute to this research program directly.

So, what's the vibe?

The Memories.ai Research Fellowship is a 12-week intensive research program. It is not an internship with assigned tasks or a shadowing arrangement. Fellows are treated as research collaborators: embedded with the team, working on open problems, and supported in producing work publishable at top-tier venues.

Here's everything you need to know.

Research Areas

Fellows work across five active research directions at Memories.ai.

Egocentric Video Understanding First-person perspective modeling from wearable cameras and AR devices. Teaching AI to understand the world as humans experience it — continuous, ego-centric, and temporally extended — rather than from fixed viewpoints or curated clips.

Large Visual Memory Models (LVMM) Memories.ai's flagship research direction. Building architectures that give AI systems persistent, retrievable visual memory across long time horizons: the ability to remember what was seen, where it was, and when — and to answer queries against that memory in real time.

Real-Time Video Intelligence Edge AI and streaming inference for live video feeds. Processing continuous video with low latency for real-world deployment in security, operations, manufacturing, and smart environments — where waiting for cloud inference is not an option.

Multimodal Memory Systems Cross-modal retrieval and reasoning across audio, video, and text. Connecting what AI sees, hears, and reads into unified memory representations that support any-to-any retrieval and multimodal RAG applications.

Data Infrastructure at Scale Scalable pipelines for processing, indexing, and retrieving massive video datasets. The engineering backbone that makes large-scale visual AI research possible — less visible than the model work, and just as important.

Who Should Apply

The fellowship is open to:

• Master's students in computer science, AI/ML, or a related field, or researchers who have recently completed a Bachelor's degree
• Researchers with hands-on experience with deep learning frameworks — PyTorch, JAX, or equivalent
• Researchers who have published or are actively working toward publication at relevant venues
• Anyone with genuine interest in video understanding, multimodal AI, egocentric vision, or large-scale data systems

FYI: Memories.ai evaluates applicants on intellectual curiosity and demonstrated research ability — not institutional pedigree. Researchers doing interesting work at any institution are encouraged to apply.

What Fellows Receive

• Unlimited cloud compute. Full access to AWS and GCP compute credits. Fellows train the models their research requires without rationing GPU hours.
• Dedicated GPU cluster access. Access to Memories.ai's infrastructure for large-scale model training and experimentation.
• Proprietary datasets. Fellows work with Memories.ai's internal video datasets and memory architectures — data not available elsewhere.
• 1:1 mentorship. Weekly one-on-one sessions with senior researchers on the Memories.ai team, including former Meta Research Scientists. Mentorship is personalized and consistent throughout the 12 weeks.
• Research network access. Connections to industry experts and academic collaborators across Memories.ai's research network.
• Publication support. Memories.ai actively supports fellows in submitting work to top-tier venues: CVPR, NeurIPS, ICLR, and others. Research produced during the fellowship is supported through the full submission process.
• $1,000/month stipend. Fellows receive a monthly stipend for the duration of the program.
• Visa sponsorship. International applicants are welcome. Memories.ai provides visa sponsorship and immigration support.

Frequently Asked Questions

What is the Memories.ai Research Fellowship? It is a 12-week paid research program for Master's students and early-career researchers in computer vision, multimodal AI, video understanding, and data systems. Fellows work directly with the Memories.ai research team on open problems in visual AI, with access to GPU infrastructure, proprietary datasets, mentorship, and publication support. 5 to 8 fellows are selected per cohort.

When does the fellowship run? The 2026 Fall Fellowship runs September through November 2026, culminating in final research presentations in November.

What is the stipend? Fellows receive $1,000 per month for the 12-week duration of the program.

Is the fellowship remote? Yes. The fellowship is remote-friendly. Optional in-person collaboration is available for fellows who want it.

Are international applicants eligible? Yes. Memories.ai provides visa sponsorship and immigration support for international applicants.

What is a Large Visual Memory Model (LVMM)? An LVMM is an AI architecture that can process, understand, and remember video at scale across long time horizons — maintaining persistent, retrievable visual memory rather than analyzing individual frames or clips in isolation. It is Memories.ai's core research direction and the system fellows contribute to during the fellowship.

What research areas can fellows work in? Fellows work in one or more of five areas: egocentric video understanding, Large Visual Memory Models, real-time video intelligence, multimodal memory systems, and data infrastructure at scale.

Do fellows get to publish their work? Yes. Memories.ai actively supports fellows in publishing at top-tier venues including CVPR, NeurIPS, and ICLR. Publication support — including co-authorship guidance, submission preparation, and research network introductions — is part of the program.

What is the cohort size? 5 to 8 fellows are selected for each cohort.

Who mentors fellows? Fellows receive weekly 1:1 mentorship from senior researchers on the Memories.ai team, including former Meta Research Scientists.

What computing resources do fellows have access to? Full access to AWS and GCP compute credits with no rationing, plus Memories.ai's dedicated GPU cluster for large-scale training and experimentation.

How do I apply? Apply via our application form and if you have any questions at all, reach out to [email protected].

Apply for the 2026 Fall Fellowship

Memories.ai is selecting 5 to 8 fellows for the Fall 2026 cohort. The program runs September through November 2026.

If you are a Master's student or early-career researcher working in computer vision, multimodal AI, video understanding, or data systems — and you want to work on problems at the frontier of visual AI with full research support — this is the program.

What's stopping you? Apply now.