AI Platform Architect

Responsibilities:
Design and own end-to-end AI platform architecture: LLM serving, RAG pipelines, agentic orchestration, and model lifecycle management.
Build scalable LLM inference infrastructure capable of serving millions of concurrent learners across diverse devices and connectivity conditions.
Define and enforce AI safety, governance, and evaluation standards across all AI systems deployed in production.
Drive RAG and agentic AI architectures: retrieval design, tool use, multi-agent frameworks, and orchestration patterns.
Lead fine-tuning pipeline design: LoRA, QLoRA, RLHF, and DPO from data preparation to production serving.
Architect MLOps practices: CI/CD for models, experiment tracking, deployment versioning, and rollback mechanisms.
Manage GPU cluster infrastructure: compute allocation, distributed training setup, and cost engineering.
Drive AI unit economics by setting up token tracking, model drift metrics, and distributed tracing via OpenTelemetry, LangSmith, or Databricks.
Collaborate with product, data, and engineering teams to translate AI platform capabilities into learner-facing features.
Mentor and grow a cross-functional team of ML engineers, platform engineers, and AI researchers.
Design and operate multi-tenant AI platform infrastructure, ensuring isolation, fairness, and cost attribution across teams.
Incorporate Indic language NLP requirements into platform design: multilingual embedding support, code-switching handling, and Indic model serving.
Requirements:
Strong experience in designing and scaling AI/LLM platforms.
Deep understanding of generative AI, LLMs, and AI system architecture.
Hands-on experience with model training, fine-tuning, and deployment (PyTorch, LoRA, RLHF, DeepSpeed, Megatron-LM).
Experience with inference frameworks such as vLLM, Triton, or TGI.
Strong knowledge of RAG, Graph-RAG, vector databases, and agent orchestration.
Understanding of distributed systems, cloud infrastructure, scalability, and security.
Experience building production-grade, multi-tenant AI applications.
Exposure to Indian-language NLP is preferred.
Deep understanding of how LLMs and foundation models are trained (not just fine-tuned) model training pipelines, distributed training, and dataset construction; hands-on experience with AI safety and security model design; proficiency in AI benchmarking methodologies and multi-agent orchestration frameworks.
Working hands-on experience with AWS (Bedrock, SageMaker), Azure AI, GCP, Docker, Kubernetes (EKS/GKE), etc. is a plus.

Experience
6-10 yrs