Fork: 140-220k net for full-time
Format: full-time or part-time (20 hours/week), remote work (including outside the Russian Federation) or office at MIPT
Description:We make MobileHealth solutions for those. giants based on neural networks. Our solutions now run on millions of devices from different companies around the world.
The tasks are non-trivial, they have a strong research component. Our task is to make a model; the customer packages it into a product.
We have curated and clean datasets + extensive R&D experience. As a result, there are few MLops and Data Engineering tasks; the main emphasis is on the quality and efficiency of the ML model.
Nevertheless, we try to keep the engineering culture high: we need to develop an infrastructure for quickly testing very different hypotheses. Bad code creates friction in research.
A project at the intersection of three areas: Computer Vision, Time-Series Analysis and Edge Computing (including model compression: quantization, pruning, distillation etc.)
What tasks will you have to do:- Code infrastructure for DL experiments in CV, Time-Series and Model Compression.
- Prepare and run reproducible and well-designed computational experiments.
- Solve engineering and research problems that arise along the way: come up with heuristics, write a custom architecture or loss, interpret non-obvious results, etc.
- Read scientific articles on the topic of the project.
- Communicate openly and closely with the team and the customer orally and in writing in Russian and English.
Knowledge, abilities, skills:- 2+ years of experience in DL, preferably CV or signals.
- Free Python, Linux, Git, work on a remote server.
- Fluent in pytorch, numpy, pandas, visualize data.
- Good mathematical base: linal, theory, matan.
- Ability to write clean and readable code.
- A creative approach, courage in ideas and a researcher’s mindset.
- Understanding of the internals of neural networks, knowledge of architectures, heuristics and metrics.
- Experience in model compression is preferred.
- Experience in publishing scientific articles is preferred.
- Management experience is preferred, one leadership position is open per project.
- Special respect for your background in digital signal processing.
- English B2+
Conditions:- Work remotely or from an office in Dolgoprudny.
- Flat structure, good opportunities for growth as a team leader.
- Possibility to work part time.
Next steps:- Leave a request on the website