Job
Description
We are looking for a skilled and creative Generative AI Engineer with 2 to 3 years of experience to join our AI/ML team. You will work on designing, fine-tuning, and deploying generative models (text, image, audio, or multi-modal) that power next-gen AI applications across business verticals. This is a hands-on role requiring a solid understanding of deep learning, model training/inference pipelines, and deployment best practices. Key Responsibilities • Build and fine-tune generative models using transformer architectures (e.g., GPT, LLaMA, T5, BERT) and diffusion models (e.g., Stable Diffusion). • Integrate LLMs into production workflows using APIs or custom backends. • Design and implement Retrieval-Augmented Generation (RAG) systems using vector databases like FAISS, Qdrant, or Pinecone. • Optimize prompt engineering and chain-of-thought logic for domain-specific tasks. • Evaluate model outputs for accuracy, relevance, hallucination rate, and latency. • Collaborate with product and engineering teams to turn prototypes into scalable systems. • Contribute to model monitoring, drift detection, and feedback loops for continual improvement. Required Skills • 23 years of experience working with machine learning models, with a focus on generative AI. • Proficient in Python and libraries such as PyTorch, TensorFlow, Transformers (Hugging Face). • Hands-on experience with OpenAI API, LangChain, or LlamaIndex. • Familiarity with vector databases (FAISS, Qdrant, Pinecone) and embedding models. • Strong understanding of LLM behavior, including prompt design, token limits, and context handling. • Experience deploying models using FastAPI, Docker, and cloud platforms (AWS, GCP, or Azure). • Good grasp of model evaluation, explainability, and safety techniques. Preferred Qualifications • Experience with multimodal AI (e.g., text-to-image, text-to-audio). • Exposure to MLOps workflows, CI/CD for ML pipelines. • Contributions to open-source AI projects or research papers. • Knowledge of privacy-preserving AI, RLHF, or federated learning is a plus.
