Indira Devi

About Indira Devi

Innovative and results-driven Generative AI Engineer with around 9 years of experience in designing, fine-tuning, and deploying large-scale AI and generative systems across domains such as healthcare, finance, telecom, and e-commerce. Highly skilled in developing Large Language Models (LLMs) like GPT, LLaMA, Falcon, and Mistral, with a strong focus on NLP, RAG, and conversational AI. Experienced in building intelligent chatbots, AI copilots, and content automation systems using LangChain, LangGraph, OpenAI, Anthropic, and vector databases like FAISS, Pinecone, and Weaviate.Proficient in deploying end-to-end AI pipelines using FastAPI, Docker, Kubernetes, and orchestrating workflows with Airflow and MLflow. Adept at cloud-native development across AWS, Azure, and GCP, ensuring scalable and fault-tolerant model serving. Deep understanding of transformer architectures and efficient training strategies including LoRA, PEFT, quantization, distillation, and distributed multi-GPU training.Hands-on expertise in image and multimodal generation using Stable Diffusion, GANs, VAEs, CLIP, Whisper, and DALL·E, applying AI for creative and enterprise use cases such as speech-to-text, captioning, and visual content generation. Experienced in safety alignment and prompt evaluation, applying RLHF, RLAIF, and feedback loops to minimize hallucination, bias, and toxicity.Strong background in data engineering and MLOps, developing RAG-powered search systems, automation pipelines, and evaluation frameworks that improve reliability and model explainability. Passionate about continuous research, keeping up with state-of-the-art innovations from NeurIPS, ICLR, and arXiv, and translating cutting-edge advancements into production-ready solutions.Recognized for mentoring teams, collaborating with cross-functional stakeholders, and transforming complex business challenges into AI-powered, scalable, and secure applications that deliver measurable value.

Experience

1. Generative Ai Engineer

   Optum

   May 2023 — Present · Eden Prairie, MN, US

   At Optum, I lead the development and deployment of cutting-edge Generative AI (GenAI) solutions, focusing on enhancing enterprise-level natural language processing (NLP) systems and intelligent automation. My core responsibility is designing, fine-tuning, and deploying Large Language Models (LLMs) for healthcare and payer-provider systems, ensuring accuracy, efficiency, and ethical AI practices.I have built robust retrieval-augmented generation (RAG) pipelines using LangChain, FAISS, Pinecone, Chroma, and OpenAI/Anthropic APIs, enabling domain-specific knowledge retrieval and real-time document understanding across Optum’s internal knowledge bases. I designed multi-turn conversational AI systems by implementing prompt chaining, contextual memory, and custom tokenization strategies, including for low-resource languages.My work includes developing prompt engineering frameworks to enable zero-shot/few-shot capabilities across summarization, Q&A, and generative tasks, while optimizing for BLEU, ROUGE, and perplexity metrics. I integrated LoRA adapters for lightweight fine-tuning and leveraged quantization (ONNX/TensorRT) and distillation/pruning techniques to deploy efficient GenAI models to mobile and edge devices with sub-100ms inference latency.Additionally, I applied reinforcement learning from human feedback (RLHF) and reward modeling to align LLM behavior with user intent. I experimented with Mixture-of-Experts (MoE) and sparse transformer architectures to reduce model training costs while maintaining performance across multilingual models like mT5, BLOOM, and XLM-R.To drive ethical AI adoption, I led red teaming exercises, designed guardrails, and enforced responsible AI practices including prompt encryption, access control, and audit-compliant logging. I partnered with security and compliance teams to ensure LLM deployments met enterprise-grade standards.

Education

• Anna University Chennai

  Bachelor of Technology (B.Tech) , Computer Science & Engineering