Harikrishnan S Nair

I’m a Generative AI Technical Lead with 9+ years of engineering experience and 3+ years specializing in enterprise Generative AI, Agentic AI systems, and AI platform engineering. I focus on delivering AI solutions that are technically deep, operationally robust, and enterprise-ready.

From GPU infrastructure provisioning to LLM deployment strategy, I lead end-to-end engineering across inference serving (vLLM, Ollama), model quantization and fine-tuning (LoRA/QLoRA), and responsible AI architecture. I build multi-agent systems and production-grade RAG pipelines that combine dense/sparse retrieval, RRF fusion, and cross-encoder reranking for measurable reliability.

At Crest AI (OMFYS’s in-house enterprise AI PaaS), I orchestrate multi-tenant, Kubernetes-native agent workloads over shared vector and graph stores (pgvector, Redis, MinIO, Neo4j) and ONNX Runtime. I’ve engineered a 18-step production RAG pipeline with 4-layer Responsible AI guardrails and real-time observability for full-stack LLMOps.

I also drive performance and cost improvements through LoRA/QLoRA (4-bit PEFT), GPU memory reduction, and GitOps CI/CD with zero-downtime deployments. My approach blends Responsible AI evaluation (PII masking, toxicity checks, NLI faithfulness), streaming analytics, and cloud-native MLOps to help teams ship safer AI that scales.