We are looking for a Generative AI Expert with strong knowledge in Retrieval- Augmented Generation (RAG) and machine learning / deep learning (ML / DL). You will work on building intelligent systems that combine large language models (LLMs) with document retrieval to generate accurate and context-aware responses.
Your role will involve developing and improving ML / DL models, fine-tuning LLMs, and integrating retrieval systems using vector databases. Youll collaborate with cross- functional teams to build real-world AI solutions that make use of both unstructured data (like PDFs and web pages) and structured sources.
Key Responsibilities :
- Design, build, and optimize RAG pipelines for document-level and multi-turn QA systems.
- Fine-tune or prompt-tune foundation models (LLMs) for domain-specific tasks.
- Develop and deploy ML / DL models to support NLP / NLU tasks like summarization, classification, and retrieval scoring.
- Integrate vector databases, semantic search tools, and embedding models for high-performance document retrieval.
- Work with unstructured and semi-structured data sources (PDFs, HTML, JSON, SQL, etc.).
- Collaborate with data engineers, ML engineers, and product teams to build end- to-end generative AI solutions.
- Monitor performance, latency, and relevance metrics; iterate on retrieval and generation models.
- Implement prompt engineering strategies and hybrid approaches (rule-based + neural) to enhance model reliability.
- Contribute to research and innovation in applied generative AI, and stay up-to- date with the latest in LLM, RAG, and MLOps ecosystems.
Key Skills Required :
Strong experience with RAG architectures and hybrid retrieval systems.Solid hands-on knowledge of LLMs (e.g., GPT, Mistral, LLaMA, Claude, DeepSeek, etc.) and embedding models (e.g., SBERT, OpenAI, HuggingFace models).Proficiency in machine learning / deep learning using PyTorch, TensorFlow, Hugging Face Transformers, etc.Experience with vector databases (e.g., FAISS, Weaviate, Pinecone, Qdrant).Experience in text chunking, retrieval scoring, prompt tuning, or LoRA / PEFT methods.Strong background in NLP, information retrieval, and knowledge graphs is a plus.Comfortable with Python and associated data science stacks (Pandas, NumPy, Scikit-learn).Experience working with real-world messy data (PDF parsing, OCR, HTML scraping, etc.)(ref : hirist.tech)