Valentin Macheret

We offer to any video provider a SaaS solution to reduce their video transcoding & encryption costs (-30%) and carbon footprint (-80%) by all means. Fast & scalable transcoding system relying on ultra low carbon infrastructure, clean source and automatic carbon compensation. Accessible to everyone via a wysiwyg user interface or a more classic REST API. >kompil, toward sustainable videos.

Research and development of an efficient video coder/decoder (codec) based on Deep Learning. We target to reduce drastically the broadcast and player carbon footprint thanks to our codec KP1; which is NPU friendly for the ongoing generations of smartphone and laptop.

► Technical stack :

- Python, Kotlin

- Pytorch (Lightning), Keras (early stage), OpenCV

- Django, Docker, GitlabCI, AWS, Scaleway, HPC, Terraform, Ansible

► AI stack :

- Deep learning, CNN, RNN, clustering

- Works on topologies and layers (activations, frequencies..)

- Enhancing learning with optimized schedulers and loss functions

- Custom quantization (int8 for ARM target)

- State of the art and frequent technology watch (AI, codec, hardware...)

Previous job duties. Moreover I tried to enhance as I could the dailylife and workflow of my teammates, clients and partners :

► Blue Frog Robotics internal developers (5-10 people) :

- Short and long term dev goals

- Setup internal best pratices / Coding style / Software craftmanship

- Code reviews, mutual help and knowledge sharing

- Continuous integration / Git flow

- Continuous deployment / build & release

- Optimization and software architecture advice

- New arrivals onboarding

► External devs / Clients :

- Enhance the homemade BlueQuark SDK

- User feedbacks

- Maintenance and support

- Organized and led training sessions with our partners

- Doc, tutorials...

► Recruitment (see previous job)