This job has expired
This position was posted on November 8, 2025 and is likely no longer accepting applications. We've kept it here for historical reference. Check out the similar jobs below!
Job Overview
Location
New York City, New York, USA
Job Type
Full-time
Category
Software Engineering
Date Posted
November 8, 2025
Full Job Description
đź“‹ Description
- • Join Reality Defender’s award-winning AI research group for a 10–12-week, fully-funded PhD internship that sits at the intersection of academic rigor and real-world impact. You will work shoulder-to-shoulder with Jacob Seidman and a cross-functional team of computer-vision scientists, ML engineers, and threat-intel analysts to push the boundaries of deepfake detection for video, image, and multimodal content.
- • Own an end-to-end research track that starts with literature review and hypothesis generation and ends with a peer-reviewed publication and a production-ready prototype. You will propose novel architectures, loss functions, or data-augmentation strategies that demonstrably improve detection accuracy against the latest generative models (e.g., Stable Diffusion XL, Sora, VASA-1).
- • Design and run large-scale experiments on our GPU cluster (A100 & H100 nodes on AWS and GCP) using PyTorch, Weights & Biases, and Slurm. You will maintain experiment reproducibility through containerized workflows (Docker + Conda) and rigorous version control (Git + DVC).
- • Curate and extend Reality Defender’s proprietary datasets—millions of real and synthetic images/videos with rich metadata—while respecting privacy and licensing constraints. You will implement active-learning pipelines to surface the most informative samples for human annotation, reducing labeling cost by up to 40 %.
- • Collaborate with our red-team to stress-test your models against adversarial attacks (e.g., GAN fingerprint removal, diffusion purification, latent perturbations). You will iterate on defense mechanisms such as watermarking, ensemble uncertainty, and transformer-based attention visualizations.
- • Contribute to our open-source evaluation suite, releasing benchmark tasks and metrics that the wider research community can adopt. Your work will directly influence the next version of the Reality Defender API, protecting elections, financial transactions, and brand reputation for Fortune-500 clients.
- • Present weekly findings to internal stakeholders and external partners (DARPA, NIST, IEEE S&P). At the end of the internship, you will deliver a conference-ready paper targeting CVPR, ICCV, or NeurIPS 2027, with full support for travel and registration.
- • Enjoy the flexibility of remote work or the energy of our NYC headquarters in SoHo—complete with standing desks, high-end GPUs on-site, and catered lunches. We maintain a “no-meeting Wednesday” policy so you can stay in flow state and maximize research velocity.
- • Receive one-on-one mentorship from senior staff who have published 100+ papers and filed 20+ patents. You will also gain exposure to product strategy, customer discovery, and VC briefings—experience rarely offered in purely academic internships.
🎯 Requirements
- • Currently enrolled in a PhD program in Computer Science, Electrical Engineering, or related field with a focus on computer vision or machine learning
- • Demonstrated research record with at least one first-author publication (or under review) at top-tier venues such as CVPR, ICCV, NeurIPS, ICML, ECCV, or TPAMI
- • Strong programming skills in Python and hands-on experience building and training deep-learning models with PyTorch; familiarity with CUDA and mixed-precision training is a plus
- • Solid understanding of convolutional and transformer architectures, generative models (GANs, VAEs, diffusion), and evaluation metrics for detection and segmentation tasks
- • Experience working with large-scale image/video datasets, data-loading pipelines (e.g., WebDataset, DALI), and distributed training on cloud GPUs
- • Availability for a full-time, 10–12-week internship during Summer 2026; U.S. work authorization is required
🏖️ Benefits
- • Competitive PhD-level stipend of $12,000–$15,000 for the summer, plus a $2,000 equipment & travel allowance
- • Flexible start and end dates that align with your academic calendar, with the option to extend into a part-time fall research assistantship
- • Direct path to publish: guaranteed co-authorship on at least one high-impact paper and full conference travel sponsorship (up to $5,000)
- • Access to Reality Defender’s proprietary datasets, compute budget (unlimited GPU hours on A100/H100 clusters), and collaboration credits on AWS & GCP
- • Mentorship from Jacob Seidman and weekly 1:1s with CTO and CEO; fast-track consideration for full-time roles post-graduation
Skills & Technologies
Python
AWS
GCP
PyTorch
Junior
Remote
Degree Required
About Reality Defender Inc.
Reality Defender develops enterprise-grade deepfake and synthetic media detection technology. Its platform analyzes images, videos, audio and text in real time to identify AI-generated tampering, phishing, fraud and disinformation for financial services, media, government and social platforms. The company provides APIs, browser extensions and on-premise solutions that integrate into existing workflows, offering risk scores, forensic evidence and continuous model updates to counter evolving generative AI threats.
