Design and build embeddings and representation learning models for traveller, property, and event data.
Develop two-tower retrieval architectures for recommendation and matching use cases.
Build models for sparse implicit-feedback scenarios and address cold-start challenges.
Design social/graph-based recommendation approaches for traveller matching and similarity.
Build retrieval and ranking pipelines, with rigorous offline evaluation practices.
Take research-backed approaches from prototype through to production deployment.
Integrate embeddings and recommendation outputs into products such as Wayfinder/recommendations.
Solid experience with embeddings and representation learning.
Hands-on experience with two-tower retrieval architectures.
Experience with sparse implicit-feedback modelling and cold-start problem-solving.




Experience with social/graph-based recommendation systems.
Proven experience building retrieval and ranking pipelines.
Solid offline evaluation methodology and rigor.
Proficiency in Python and up-to-date ML frameworks.
Demonstrated research-to-production experience, not research-only.
Likely scope includes traveller, property, and event embeddings; social similarity; event relevance; and integration into recommendation systems.
Avoid: generic GenAI/NLP profiles without recommendation systems experience.
Screening question: "Given stays, co-location, profile views, chats, Linkups and event attendance, how would you build and evaluate traveller embeddings?"

📌 Senior Applied Ml Scientist Bengaluru
🏢 Tavant Technologies
📍 Bengaluru