Akash Thakur

I’m a Sr. AI Architect with 10+ years of experience building and scaling production-grade AI systems. I specialize in Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and multi-agent architectures—delivering end-to-end compliance-ready AI platforms for high-volume enterprise workloads with minimal inference latency, cost optimization, and complete automation.

In my recent work, I focus heavily on LLMOps, MLOps, and model optimization. I bring hands-on experience in fine-tuning, prompt engineering, and scalable model serving so teams can deploy reliably in real production environments.

Earlier, as a Senior AI Technical Lead, I built an autonomous agentic platform for root-cause analysis of application failures. The system eliminated manual log interpretation across distributed systems, reducing manual effort by 68% and MTTR by 82%, using a multi-stage agentic pipeline with LangGraph-based orchestration, LLM reasoning, retrieval-augmented context, and multi-source data processing.

I’ve also delivered real-time AI outcomes through prompt engineering, quantization, and vector-based retrieval for low-latency processing (including Pinecone). In parallel, I designed RAG-based multi-agent systems to retrieve precise answers from 15+ GB of Confluence and runbooks, refined query understanding using Transformer Models in PyTorch, and optimized retrieval with FAISS—improving productivity by 39% and reducing IT/helpdesk query load by 43%; I also contributed an AI patent/OCR solution using transformers, OpenCV, and XGboost for auto-template classification and extraction (99% accuracy printed, 82% handwritten) and share details in my article.