Senior Data Scientist, Recommendation System

Position : Senior Data Scientist, Recommendation Systems 

This position open for Principal level too, depend on depth of experience, leadership in methodological innovation and impact in production systems.

Location: Open globally (strong overlap with GMT+3 to GMT+8 preferred)

About Role 

We're looking for a Senior Data Scientist, Recommendation Systems to own and orchestrate our recommendation infrastructure across all surfaces. You'll design how multiple recommendation models work together, balance competing constraints (user preferences, pharma content, content freshness, diversity), and build the semantic search and embedding layer that powers personalized experiences for doctors. 

This isn't just about building another recommendation algorithm, it's about architecting how collaborative filtering, user-based filtering, embeddings, and business constraints combine into a unified system that serves the right content to the right doctor at the right time. You'll work at the intersection of recommendation algorithms, constraint optimization, vector databases, and production systems. 

You'll be the person who ensures our recommendation strategy is coherent, scalable, and balances user value with business objectives. 

What You’ll Do 

Recommendation Strategy & Orchestration (40%) 

• Design overall recommendation architecture across all surfaces (Feed, Search, Swipe, Network Recommendation) 
• Orchestrate how multiple models work together built by other DS. 
• Build multi-objective optimization frameworks that balance : user preferences, business constraints and rules, content freshness, diversity and other attributes required. 
• Design ranking and retrieval strategies 
• Define recommendation quality metrics (relevance, diversity, novelty, coverage) 
• Work with other DS to A/B test recommendation strategies 

Deep Analysis & Optimization (40%) 

• Analyze recommendation system performance across surfaces 
• Measure quality metrics: click-through rate, engagement, diversity, coverage, serendipity 
• Identify failure modes: why do recommendations fail? What patterns are we missing? 
• Apply statistical methods to understand recommendation effectiveness 
• Debug recommendation issues (filter bubbles, cold start, popularity bias) 
• Optimize constraint satisfaction (linear programming, Pareto optimization) 
• Work with Behavioral DS to incorporate behavioral signals into ranking 
• Collaborate with existing DS on model improvements 

Vector Database & Embedding Infrastructure (20%) 

• Build and maintain vector database schema for semantic search and retrieval. You will work with Data Architect, Data Engineer and DevOps to productionize your work and infrastructure stuff. 
• Create and optimize embedding strategies for: medical content, doctor profiles and constraint’s campaigns 
• Implement retrieval systems 
• Fine-tune embedding models for medical domain (or work with NLP DS for custom models) 
• Optimize for latency and scale (serving recommendations to thousands of doctors) 
• Design embedding versioning and retraining pipelines 

Cross-Functional Collaboration 

• With Junior DS: Provide technical direction on how their models integrate into the overall system 
• With Behavioral Science DS: Incorporate behavioral features into recommendation ranking 
• With Experimentation DS: Design A/B tests for recommendation strategies 
• With NLP DS: Collaborate on embedding quality for medical content 
• With Data Engineers & Data Architects: Design vector database infrastructure and embedding pipelines 

Who You Are 

We welcome candidates with deep expertise in architecting and orchestrating production recommendation systems at scale. You are a technical leader who thinks in terms of systems, trade-offs, and infrastructure, not just algorithms. 

Minimum: Bachelor's degree in Computer Science, Statistics, Mathematics, Data Science, Physics, or relevant field 

Preferred: Master's or PhD 

Years of Experience: Min. 6 years in recommendation systems, ML engineering, or related fields 

What matters most: Deep understanding of recommendation algorithms, experience with embeddings and vector databases, ability to architect systems that balance multiple objectives, and production ML experience 

Core Recommendation Systems Expertise (Required Methodological Foundation) 

Required: Deep, hands-on experience in the following pillars of modern recommendation systems: 

• Recommender System Architecture & Orchestration: 

1. Designing hybrid recommendation systems that combine collaborative filtering, content-based, and embedding-based approaches produced by other DS. 
2. Building multi-stage retrieval & ranking systems (e.g., candidate generation, ranking, re-ranking). 
3. Solving cold-start and long-tail challenges. 

• Constraint Optimization & Ranking: 

1. Formulating and solving multi-objective optimization problems (e.g., linear programming, Pareto frontiers). 
2. Designing ranking functions that balance user preference, business rules, diversity, and freshness. 

• Evaluation & Analysis: 

1. Expertise in recommendation metrics: NDCG, MAP, diversity, coverage, serendipity. 
2. Designing and analyzing A/B tests for recommendation strategies. 
3. Diagnosing systemic issues: filter bubbles, popularity bias, feedback loops. 

• Embedding Infrastructure & Semantic Search: 

1. Building and optimizing embedding models (two-tower, sentence transformers) and vector search systems. 
2. Hands-on experience with vector databases (at least FAISS or Pinecone, Weaviate, Qdrant) and approximate nearest neighbor (ANN) algorithms. 
3. Fine-tuning embeddings for domain-specific performance. 

Essential Technical Execution Skills 

• Programming & ML Ops: Expert proficiency in Python and its ML ecosystem (e.g., PyTorch/TensorFlow, scikit-learn). A track record of writing clean, tested, production-ready code. 
• Data & Infrastructure: Advanced SQL for complex data analysis. Proficiency with Git and collaborative development workflows. 
• Systems Understanding: Experience with large-scale low-latency serving architectures (real-time and batch) and familiar with MLOps practices (model versioning, monitoring, pipelines). 

Valuable Domain Knowledge (Significant bonus) 

• Experience in healthcare, medical, or other highly regulated content domains. 
• Understanding of healthcare professional workflows, learning, or decision-making. 
• Familiarity with medical ontologies or domain-specific embeddings (e.g., BioBERT, ClinicalBERT). 

Advanced & Cutting-Edge Skills (Bonus / Growth Areas) 

Experience in any of these is a strong plus: 

• Transformer-based or Graph Neural Network (GNN) recommendation models. 
• Reinforcement Learning or Contextual Bandits for adaptive recommendation. 
• Causal inference methods to evaluate recommendation impact.