Job Description
We are working with a company who is building the best business AI video system on the market. Powered by the next-generation video artificial intelligence, they deliver unprecedented insights and 10x better user experience than the incumbents of the vast but stagnant video security industry.
The company is a tech start up which is well funded and has been going for 3 years now.
The role
We are hiring a Machine Learning engineer.
* Take an existing open-source Pytorch model, fine tune, productionize them in C++ runtime, and optimize for latency and throughput.
* Take an open-source model and fine-tune them on our in-house data set as needed.
* Design thoughtful experiments in evaluating the tradeoffs between latency and accuracy on the end customer use case.
* Integrate the model with the downstream use case and fully own the end metrics.
* Maintain and improve all existing ML applications in the product
* Read research papers and develop ideas on how they could be applied to video security use cases, and convert those ideas to working code.
Requirements
* You should be a good software engineer who enjoys writing production-grade software.
* Strong machine learning fundamentals (linear algebra, probability and statistics, supervised and self-supervised learning)
* Keeping up with the latest in deep learning research, reading research papers, and familiarity with the latest developments in foundation models and LLMs
* (Good to have) Comfortable with productionzing a Pytorch model developed in C++, profiling the model for latency, finding bottlenecks, and optimizing them
* Good understanding of docker and containerization
* (Good to have) experience with Pytorch and Python3, and comfortable with C++
* (Good to have) Understanding of Torch script, ONNX runtime, TensorRT
* (Good to have) Understanding of half-precision inference and int8 quantization