about
hi, i'm vlad kostoglodov.
you can call me vkost.
I work on recommendation systems, from representing interaction data to retrieving and ranking candidates.
The open work here follows a few threads: sparse linear models, sequential signals, and the practical limits of training and inference.
selected work
sparse models / performanceSANSA / same weights, less waiting ↗
Reuse residual columns and trim sparse computation while keeping the resulting weights unchanged.
2.04× faster fit · 17–18% less peak RAM on two datasetsretrieval / ranking / team solutionAvito ML Cup / one matrix, several histories ↗
Reuse learned relationships with different views of user history, then rank forty catalogue nodes.
1st place · team AmazMe · 2025cold start / constrained ranking / team solutionVK LSVD / new video, familiar author ↗
Find an audience through author history, then allocate users across videos under a shared budget.
3rd place · Elves do RePlay · 2025A few more threads: linear autoencoders, stateful inference in Riiid, and retrieval evidence in music dialogues.
Kaggle Competitions Expert. More solutions in competitions.