* Manage 10+ engineers for Home Shortcuts & Now playing View, driving organic recommendations strategy

* Launched Ray + PyTorch Transformer models for familiar recommendations, lifting Weekly Active Users

* Improved Now Playing View ranking to make it the primary driver for Live Events & Artist Connections

* Engineer conversational interaction models for Spotify iDJ using SFT on internal LLMs

* Led 6 engineers to personalize ad targeting, reducing Podcast ads cost-per-stream (CPS) by ~24% using MMoE architecture

* Leveraged LLM-based embeddings to increase video ad impressions by 7x for newly onboarded creators‬

* Conducted meta-analysis of ranking experiments to reduce the user sample required for A/B testing by 43%

* Formed cross-functional team (ML, data, backend) to drive organic recommendations for long-term retention

* Developed RL reward models to optimize Home & Search recommendations, boosting podcast discovery by 29%

* Built offline simulator to predict retention impact of recommendations; published in KDD 2024, ForTune

* Built LTV models using Beta-Geometric Survival & Markov Chains; metrics reported in Spotify Investor Calls

* Optimized Google Ad upsell targeting, driving 0.61% retention lift and 5.5% increase in Premium conversion

* Developed a new personalization framework to generate “in-feed” account recommendations for users’ interests.

* Built calibrated account recommendation models for 1B+ users, providing core signals to real-time Ads bidding

* Scaled account recommendation delivery by 120% via impression discounting and caching strategies