LLM/ML/NLP - Engineer - Pan India

Key Responsibilities:

Machine Learning (ML):

• Design, build, and deploy

  ML pipelines

  for structured and unstructured data.
• Conduct

  feature engineering

  , model selection, and hyperparameter tuning.
• Evaluate model performance using industry-standard metrics and improve accuracy.
  
• Collaborate with data engineering teams to ensure clean and accessible data for model training.
  
• Apply supervised, unsupervised, and reinforcement learning techniques as per use case.
  

Natural Language Processing (NLP):

• Develop NLP pipelines for

  text classification, sentiment analysis, entity recognition, summarization

  , etc.
• Leverage libraries like

  spaCy, NLTK, Hugging Face Transformers, and Gensim

  .
• Preprocess and tokenize large corpora using advanced NLP methods.
  
• Implement solutions for multi-lingual, domain-specific text data challenges.
  
• Integrate NLP services with applications or workflows.
  

Large Language Models (LLM):

• Fine-tune and deploy LLMs (e.g.,

  GPT, LLaMA, BERT, Falcon, Mistral

  ) on custom datasets.
• Use

  prompt engineering

  and

  retrieval augmented generation (RAG)

  to build intelligent systems.
• Optimize inference performance and latency for production-level deployment.
  
• Stay updated with the latest in GenAI, foundation models, and transformer architectures.
  
• Work on use cases like

  chatbots, question answering, summarization, content generation

  , and

  semantic search

  .

Required Skills:

• Proficiency in

  Python

  ,

  PyTorch

  or

  TensorFlow

  , and data science libraries (

  scikit-learn

  ,

  pandas

  ,

  NumPy

  ).
• Experience with

  NLP frameworks

  and

  transformer-based architectures

  .
• Exposure to

  LLM model training/fine-tuning

  using tools like Hugging Face, LangChain, or OpenAI API.
• Familiarity with

  MLOps

  , model versioning, and deployment using Docker/Kubernetes.
• Understanding of

  vector databases

  (e.g., FAISS, Pinecone, Weaviate) and embeddings.