Video Research Intern (6months internship)

MediaKind; Rennes; Hybrid; EUR 13,000 - 17,000; Full time

Benefits offered by this job

Paid internship

Public transport reimbursement 50%

Meals 50% meal allowance

Job summary

MediaKind is offering a six-month Master’s-level internship in Rennes, France, focused on applied video research. You will work under close supervision while owning a well-defined piece of work, from literature review to experimentation and a shareable result.

You will set up experiments with codecs and datasets, build reproducible evaluation tools, and contribute to technical reports and internal presentations with mentorship from senior researchers.

Qualifications

• Hands-on experience in machine learning or deep learning, ideally on image or video problems, using Python and a framework such as PyTorch or TensorFlow.
• Understanding of image/video processing and compression fundamentals (e.g. filtering, transform coding, motion compensation, rate-distortion).
• Ability to read, critically assess and summarize research papers.
• Clear written and spoken English (French is a plus).

Responsibilities

• Run structured literature and state-of-the-art reviews and summarize findings for the team.
• Set up and run experiments with open-source or proprietary codecs, models and datasets.
• Build and maintain reproducible evaluation, benchmarking and experimentation tools.
• Investigate, develop and test proofs of concept (PoCs) with guidance from senior researchers.
• Analyze results and document them in technical reports and team presentations.
• Contribute to code deliveries, publications or patent disclosures.

Skills

ML/DL

Python

PyTorch

TensorFlow

Literature review

Education

Master's student in Electrical Engineering/CS/Video Processing

Tools

FFmpeg

x264/x265

SVT-AV1

VVenC

VTM

Job description

Who we are

At MediaKind, we’ve reshaped the future of streaming video.

As the world’s largest independent video technology company, and the leader in sports streaming, we power the most ambitious and demanding video experiences on the planet.

Our cloud-native, software-defined solutions help the world’s leading streamers and broadcasters build, deploy and scale premium video experiences. From live sport and entertainment to mission-critical broadcast, we help them deliver every moment to audiences, everywhere.

At MediaKind, innovation is driven by people. You'll join a global team of curious thinkers, problem-solvers, and builders who move quickly, collaborate openly, and are passionate about delivering technology that makes a real impact. We value ownership, practical thinking, and continuous learning, and we believe great ideas come from empowered teams.

If you're excited by solving complex challenges, working alongside talented colleagues around the world, and helping shape the future of video, we'd love to build what’s next together.

About the Team

MediaKind provides cloud-native video delivery and streaming solutions to broadcasters and service providers worldwide. Our Applied Video Research team is a small, highly focused group that turns advances in video compression, perceptual quality and machine learning into practical improvements for MediaKind products, such as encoding efficiency, quality monitoring and viewer experience.

About the Internship

As a Video Research Intern, you will actively support our applied video research roadmap. You will work under the close supervision of senior team members, who guide your scope, methods and reviews. You will own a well-defined piece of work, from state-of-the-art review through experimentation to a clear, shareable result. This is a Master's-level internship for a duration of 6 months (ideally starting in January / February 2027) and based in Rennes, France on a hybrid working mode.

Research Topics

Your project will be defined with you, according to your background and interests, around one or more of these areas (you are not expected to know them all in advance):

• End-to-end learned compression: Survey the state of the art in neural video coding and assess practicality and feasibility against standard codecs (e.g. AVC, HEVC, AV1, VVC) under realistic deployment conditions, covering coding efficiency, complexity, latency and deployability.

• Deep-learned perceptual quality metrics: Survey and benchmark the state of the art in full-reference video quality metrics (e.g. PSNR, SSIM, VMAF and learned metrics), for customer QoE monitoring and encoder tuning. Analyze their limits (e.g. alignment with human perception, local degradations hidden by global averaging, computational cost) and explore perception-aligned, local-aware and low-complexity improvements.

• Perceptual and saliency maps: Review existing approaches (e.g. visual saliency and semantic analysis of perceptually sensitive regions such as faces, text and moving objects) and help design maps that can drive objective metrics or codec features, such as adaptive quantization and bit allocation.

• Neural pre-/post-processing: Review and optimize neural processing stages around standard codecs (e.g. resampling, denoising, enhancement) and evaluate their impact on rate-distortion and perceived quality.

What You’ll Do

• Run structured literature and state-of-the-art reviews and summarize findings for the team.

• Set up and run experiments with open-source or proprietary codecs, models and datasets, depending on the topic.

• Build and maintain reproducible evaluation, benchmarking and experimentation tools.

• Investigate, develop and test proofs of concept (PoCs) with guidance from senior researchers.

• Analyze results and document them in technical reports and team presentations, sharing knowledge across the team.

• Contribute, where appropriate, to code deliveries, publications or patent disclosures.

Qualifications

Eligibility

• Enrolled in a Master’s program or the final year of an engineering degree, in Electrical Engineering, Computer Science, Image/Signal Processing, Applied Mathematics, Machine Learning, Telecommunications, Multimedia or a related field.

• Able to sign an internship agreement (convention de stage) with your school or university, which is mandatory for this position.

Required

• Hands-on experience in machine learning or deep learning, ideally on image or video problems, using Python and a framework such as PyTorch or TensorFlow.

• Understanding of image/video processing and compression fundamentals (e.g. filtering, transform coding, motion compensation, rate-distortion).

• Ability to read, critically assess and summarize research papers.

• Clear written and spoken English (French is a plus).

Preferred (any of the following is a plus)

• Hands-on experience with video codecs, reference models and tools (e.g. x264/x265, SVT-AV1, VVenC, VTM, ECM or FFmpeg).
• Familiarity with learned image/video compression.
• Knowledge of objective and subjective video quality assessment.
• C/C++ programming experience.
• Familiarity with Linux and Git.

Don’t tick every box? If the role excites you and you meet most of the requirements, we encourage you to apply.

What You’ll Gain

• Hands-on experience applying state-of-the-art research in a commercial video streaming context.
• Access to both open-source and industrial-grade proprietary video compression or processing technology.
• Regular mentoring from experienced video researchers.
• Exposure to the full path from academic literature to PoC to product-relevant evaluation.
• Opportunity to present your work and contribute to internal reports, codebase, and potentially external publications or patents.

Benefits

• Paid internship
• Public Transport: 50% reimbursement of your monthly transit pass (e.g. STAR network in Rennes)
• Meals: 50% daily meal allowance via meal vouchers (tickers restaurant).

Here at MediaKind, we are dedicated to creating an inclusive workplace where we appreciate the unique skills, capabilities and perspectives that our people bring. We believe that diverse teams are better for our employees, customers and our business. Everyone’s identity, background and life experiences add to the MediaKind story, and we welcome all.