Sergei Bogdanov

• Designed and implemented robust algorithm for absolute daily searches estimation for the given topics. • Responsible for design, BackEnd, ML & DevOps. • Created scalable version of this algorithm by utilising AWS lambda functions. • Decreased prediction fluctuation on noisy data by 3 times. • Deployed project on AWS and developed API: EC2, lambda, RDS, S3, API Gateway, VPC, ECS

• Developed quality assurance concurrent automatic tests of our products for dev & prod environments. • Responsible for gRPC API expansion and Machine Learning models deployment. • Stack: Go, Python, Docker, Kubernetes, Redis, PostgreSQL, S3.

• Solved Classification, Regression, Object Detection, Object Recognition, Model Compression, Segmentation, Image Reconstruction (AE, VAE), Best Frame Selection problems. • Increased Face Detection MAP from 0.68 to 0.82 on Masked Faces. • Decreased Face Recognition RAM usage (now 3 times smaller) without quality decrease. • Increased Face Identification TPR from 0.09 to 0.99 with FPR=1e-3 on Masked Faces. • Increased Face Identification TPR from 0.94 to 0.96 with FPR=1e-7 on company’s private dataset.

• Developed model for Method name generation task which is comparable to SOTA • Developed model for Code-SearchNet task: Top-20% out of all competition participants.