AI scientist

Responsibilities:

• Architect and lead the development of AI models that integrate

  multimodal data (text + image)

  for product classification, attribute extraction, and vision-text consistency validation.
• Design and evolve

  taxonomy intelligence frameworks

  , including canonical entity graphs, category hierarchies, and attribute contracts for retail product data.
• Build and optimize

  CLIP-style embedding models

  to assess image-category similarity and detect mismatches between product visuals and descriptions.
• Design scalable

  LLM-based pipelines

  for zero/few-shot classification, brand normalization, and policy enforcement across diverse product types.
• Lead the creation of

  graph-based reasoning systems

  for alias resolution, brand variants, and hierarchical relationships (e.g., Yoga Lenovo Yoga”).
• Collaborate with data engineers, search relevance teams, and business stakeholders to ensure taxonomy and multimodal intelligence are tightly integrated into the search and ranking pipeline.
• Guide the development of

  data quality scoring systems

  using multimodal confidence metrics and taxonomy correctness.
• Mentor a team of AI scientists and engineers, providing technical leadership, code reviews, and strategic direction.
• Drive experimentation and active learning strategies to continuously improve model performance and reduce manual review load.
• Ensure scalability and cost-efficiency through model distillation, caching, and tiered inference strategies.

Qualifications:

• Degree in

  Computer Science, Machine Learning, Computational Linguistics, or a related field

  .

• 7+ years of experience

  in applied AI/ML, with a strong focus on

  multimodal learning

  ,

  taxonomy modeling

  , or

  knowledge graphs

  .
• Deep expertise in

  LLMs (e.g., GPT, T5, LLaMA)

  and

  VLMs (e.g., CLIP, BLIP, Flamingo)

  for classification, embedding generation, and cross-modal alignment.
• Proficiency in

  Python

  and ML frameworks such as

  PyTorch, TensorFlow, Hugging Face Transformers, OpenCLIP

  .
• Hands-on experience with

  graph databases

  (e.g., Neo4j, Amazon Neptune) and

  graph algorithms

  for entity resolution and taxonomy mapping.
• Strong understanding of

  semantic similarity

  ,

  embedding-based retrieval

  , and

  zero-shot learning

  techniques.
• Good to have - experience deploying models in production using

  MLOps tools

  (e.g., MLflow, Kubeflow, SageMaker).
• Familiarity with

  retail product data

  , e-commerce taxonomies, and consumer search behavior.
• Proven leadership in guiding cross-functional teams and mentoring junior scientists.
• Excellent communication and stakeholder engagement skills, with the ability to translate complex AI concepts into business impact.