Principal AI/ML Engineer

Optum India; Bengaluru; On-site; INR 4,000,000 - 8,000,000; Full time

Job summary

Optum India in Bengaluru seeks a Principal Data Scientist ML/DL to drive state-of-the-art AI solutions for medical applications and lead end-to-end research to production workflows. You will advance NLP technologies, information extraction, and cloud-based deployments, translating research into scalable clinical AI products.

The role demands a PhD or MS with 12–15+ years in applied AI/ML, with strong publication records and a deep fluency in Python, PyTorch, NLP, and transformer architectures.

Qualifications

• PhD or master’s degree in computer science, Machine Learning, or related field.
• 12-15+ years of experience in applied AI/ML with statistics, with a strong track record of delivering production-grade models.
• Deep expertise in: NLP, Fundamental machine learning, deep learning, transformer, state space-based architecture.
• Strong in Python coding, SQL and database queries, data preparation, and analysis
• Exploratory Data Analysis (EDA)
• Experience with PyTorch
• LLM training and fine tuning (e.g., GPT, LLaMA, Mistral, Qwen)
• Graph-based retrieval systems (knowledge graphs)
• Semantic search and vector databases (e.g., FAISS, Weaviate, Milvus)
• Model fusion and ensemble techniques (stacking, boosting, gating)
• Optimization algorithms (Bayesian, Particle Swarm, Genetic Algorithms)
• Reinforcement learning (e.g., RLHF, PPO, DPO, GRPO), Supervised Fine Tuning (SFT), LoRA, QLoRA, axolotl
• Bonus Skills
• Experience with healthcare data and medical coding systems (e.g., CPT, CM, PCS).
• Familiarity with regulatory and compliance frameworks in AI deployment.
• Contributions to open-source AI projects or published research. And/Or ability to take research papers to poc – production.
• Document segmentation and preprocessing (OCR, layout parsing)
• Distributed training frameworks (NCCL, Horovod, DeepSpeed)
• High-performance networking (InfiniBand, RDMA)
• Symbolic AI and rule-based systems
• Meta-learning and Mixture of Experts architectures

Responsibilities

• Lead end-to-end training and fine-tuning of Large Language Models (LLMs), including both open-source (e.g., Qwen, LLaMA, Mistral) and closed-source (e.g., OpenAI, Gemini, Anthropic) ecosystems.
• Deep knowledge and extensive experience with Machine/Deep Learning frameworks including transformer architectures, state space models, large language models, and agentic approaches
• Knowledge of algorithms and techniques within a computational domain with emphasis on text processing
• Demonstrated publication record in AI domain especially relating to text extraction and summarization
• Architect and implement GraphRAG pipelines, including knowledge graph representation and retrieval for enhanced contextual grounding.
• Design, train, and optimize semantic and dense vector embeddings for document understanding, search, and retrieval.
• Develop semantic retrieval systems with advanced document segmentation and indexing strategies.
• Build and scale distributed training environments using NCCL and InfiniBand for multi-GPU and multi-node training.
• Apply reinforcement learning techniques (e.g., RLHF, RLAIF) to align model behavior with human preferences and domain-specific goals.
• Experience with Hybrid NLP solutions that combine symbolic and machine learning approaches
• Collaborate with cross-functional teams to translate business needs into AI-driven solutions and deploy them in production environments.

Skills

Python

SQL

PyTorch

NLP

Deep learning

Transformer architectures

Knowledge graph retrieval

Education

PhD or Master in CS/ML

Tools

FAISS

Weaviate

Milvus

NCCL

DeepSpeed

Job description

The Principal Data Scientist ML/DL is a primary driver in the design and development of state-of-the-art Artificial Intelligence solutions for medical applications. The Principal Data Scientist ML/DL works closely with senior data scientists, machine learning engineers, software engineers and subject matter experts on current company technologies and forward-looking projects. They are the drivers of new research and solution implementation in the creation of novel artificial intelligence approaches.

Primary responsibilities include the enhancement of existing company NLP technologies and extension of those systems in new cloud-based applications. Emphasis is on development of novel machine/deep learning techniques for information extraction and synthesis. They translate research code into clinical NLP solutions deployed at scale in production environments including statistical methods, deep learning, and large language model technologies. Work will involve all aspects of methods development from initial PoC implementation to performance characterization and production launch of new methods.

The successful candidate will have a strong history of publication in Machine/Deep Learning with an emphasis on Natural Language Processing, Information Retrieval and/or Information Extraction. Exposure to recent research literature and the ability to effectively implement new technologies is key. The successful candidate will have proven success in taking machine/deep learning solutions to production environments. Strong technical skills are required.

Key Responsibilities

• Lead end-to-end training and fine-tuning of Large Language Models (LLMs), including both open-source (e.g., Qwen, LLaMA, Mistral) and closed-source (e.g., OpenAI, Gemini, Anthropic) ecosystems.
• Deep knowledge and extensive experience with Machine/Deep Learning frameworks including transformer architectures, state space models, large language models, and agentic approaches
• Knowledge of algorithms and techniques within a computational domain with emphasis on text processing
• Demonstrated publication record in AI domain especially relating to text extraction and summarization
• Architect and implement GraphRAG pipelines, including knowledge graph representation and retrieval for enhanced contextual grounding.
• Design, train, and optimize semantic and dense vector embeddings for document understanding, search, and retrieval.
• Develop semantic retrieval systems with advanced document segmentation and indexing strategies.
• Build and scale distributed training environments using NCCL and InfiniBand for multi-GPU and multi-node training.
• Apply reinforcement learning techniques (e.g., RLHF, RLAIF) to align model behavior with human preferences and domain-specific goals.
• Experience with Hybrid NLP solutions that combine symbolic and machine learning approaches
• Collaborate with cross-functional teams to translate business needs into AI-driven solutions and deploy them in production environments.

Required Qualifications

PhD or master’s degree in computer science, Machine Learning, or related field.

12-15+ years of experience in applied AI/ML with statistics, with a strong track record of delivering production-grade models.

Deep expertise in: NLP, Fundamental machine learning, deep learning, transformer, state space-based architecture

Strong in Python coding, SQL and database queries, data preparation, and analysis

Exploratory Data Analysis (EDA)

Experience with PyTorch

LLM training and fine-tuning (e.g., GPT, LLaMA, Mistral, Qwen)

Graph-based retrieval systems (knowledge graphs)

Semantic search and vector databases (e.g., FAISS, Weaviate, Milvus)

Model fusion and ensemble techniques (stacking, boosting, gating)

Optimization algorithms (Bayesian, Particle Swarm, Genetic Algorithms)

Reinforcement learning (e.g., RLHF, PPO, DPO, GRPO), Supervised Fine Tuning (SFT), LoRA, QLoRA, axolotl

Bonus Skills

Experience with healthcare data and medical coding systems (e.g., CPT, CM, PCS).

Familiarity with regulatory and compliance frameworks in AI deployment.

Contributions to open-source AI projects or published research. And/Or ability to take research papers to poc – production.

Document segmentation and preprocessing (OCR, layout parsing)

Distributed training frameworks (NCCL, Horovod, DeepSpeed)

High-performance networking (InfiniBand, RDMA)

Symbolic AI and rule-based systems

Meta-learning and Mixture of Experts architectures