Data Scientist

Data Scientist

Key Responsibilities:

• Develop, train, and deploy recommendation models using

  two-tower, multi-tower, and cross-encoder architectures

  .
• Generate and utilize text/image embeddings (e.g.,

  CLIP

  ,

  BERT

  ,

  Sentence Transformers

  ) for content-based recommendations.
• Design

  semantic similarity search pipelines

  using vector databases (FAISS, ScaNN, Qdrant, Matching Engine).
• Create and manage scalable ML pipelines using

  Vertex AI

  ,

  Kubeflow Pipelines

  , and

  GKE

  .
• Handle large-scale data preparation and feature engineering using

  Dataproc (PySpark)

  and

  Dataflow

  .
• Implement cold-start strategies leveraging metadata and multimodal embeddings.
• Work on

  user modeling

  ,

  temporal personalization

  , and

  re-ranking strategies

  .
• Run A/B tests and interpret results to measure real-world impact.
• Collaborate with cross-functional teams (Engineering, Product, DevOps) for model deployment and monitoring.

Must-Have Skills:

• Strong command of

  Python

  and ML libraries: pandas, polars, numpy, scikit-learn, matplotlib, tensorflow, torch, transformers.
• Deep understanding of

  modern recommender systems

  and

  embedding-based retrieval

  .
• Experience with

  TensorFlow

  ,

  Keras

  , or

  PyTorch

  for building deep learning models.
• Hands-on with

  semantic search

  ,

  ANN search

  , and

  real-time vector matching

  .
• Proven experience with

  Vertex AI

  ,

  Kubeflow on GKE

  , and ML pipeline orchestration.
• Familiarity with vector DBs such as

  Qdrant

  ,

  FAISS

  ,

  ScaNN

  , or

  Matching Engine

  on GCP.
• Experience in deploying models via

  Vertex AI Online Prediction

  ,

  TF Serving

  , or

  Cloud Run

  .
• Knowledge of

  feature stores

  ,

  embedding versioning

  , and

  MLOps practices

  (CI/CD, monitoring).

Preferred / Good to Have:

• Experience with

  ranking models

  (e.g.,

  XGBoost

  ,

  LightGBM

  ,

  DLRM

  ) for candidate scoring.
• Exposure to

  LLM-powered personalization

  or hybrid retrieval systems.
• Familiarity with

  streaming pipelines

  using

  Pub/Sub

  ,

  Dataflow

  ,

  Cloud Functions

  .
• Hands-on with

  multi-modal retrieval

  (text + image + tabular data).
• Strong grasp of

  cold-start problem solving

  , using enriched metadata and embeddings.

GCP Stack You’ll Work With:

• ML & Pipelines:

  Vertex AI, Vertex Pipelines, Kubeflow on GKE

• Embedding & Retrieval:

  Matching Engine, Qdrant, FAISS, ScaNN, Milvus

• Processing:

  Dataproc (PySpark), Dataflow

• Ingestion & Serving:

  Pub/Sub, Cloud Functions, Cloud Run, TF Serving

• CI/CD & Automation:

  GitHub Actions, GitLab CI, Terraform