Abhishek Rath
Vice President Agentic Ai & Llm Platforms @Wells Fargo
Signup · Get unlimited contacts
WORK HISTORY
Vice President Agentic Ai & Llm Platforms @Wells Fargo
Hyderabad, IN
Multi-Agent Architecture: Architected enterprise-wide multi-agent orchestration platform using LangGraph with Master Router pattern coordinating 15+ specialized sub-agents, achieving 98% uptime in production financial workflows• Architected comprehensive drift monitoring system for production LLM-based multi-agent architecture serving financial services workflows• Designed and deployed production RAG systems using hybrid retrieval (dense semantic + sparse BM25) with cross-encoder re-ranking, improving answer accuracy from 72% to 94%• Cost Architecture & TCO Optimization: Engineered a holistic cost framework that delivered $2.5M+ in annual TCO savings. • Architectural Authority & Governance: Served as the primary technical authority for a 50+ engineer organization, establishing architectural governance and standardization. Defined reusable agent design patterns (such as ReAct, Plan-and-Execute, and RAG) to balance developer productivity with production scalability, guiding build vs. buy decisions for agent infrastructure
EDUCATION
Cv Raman college of engineering
Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering
ABOUT ABHISHEK RATH
I am an Agentic AI Architect and specialize in designing and deploying production-grade agentic AI and multi-agent systems, with over 12 years of experience building scalable, reliable AI platforms for mission-critical environments.My core expertise lies in architecting autonomous multi-agent systems using LangGraph (planner–executor patterns, master router architectures, and ReAct-style tool-using agents), MCP-based standardized tool integration, and Agent-to-Agent (A2A) communication protocols. I focus on treating large language models as controlled black boxes layering prompt constraints, retrieval grounding, validation, and observability to ensure predictable, enterprise-safe behaviour.In my current role as Vice President - Agentic AI & LLM Platforms at Wells Fargo, I architected an enterprise-wide multi-agent orchestration platform coordinating more than 15 specialized agents, achieving 98% production uptime across financial workflows. I also led the design of hybrid RAG systems with dense and sparse retrieval, cross-encoder re-ranking, and rigorous evaluation pipelines, improving answer accuracy from 72% to 94% while delivering over $2.5M in annual cost savings through architectural optimization.A central focus of my work has been long-running agent reliability. I am the author of an arXiv-published research paper on Agent Drift in multi-agent LLM systems, where I introduced a 12-metric framework spanning semantic, coordination, and behavioural degradation, along with a composite Agent Stability Index (ASI) for production monitoring. This research directly informs how I design agent observability, drift detection, and mitigation strategies in real-world systems.What distinguishes my approach is a strong bias toward production realism: evaluation before deployment (RAGAS, LLM-as-a-Judge), confidence thresholding and refusal mechanisms, policy enforcement, and architectural governance that balances developer velocity with system safety. I am deeply motivated by building agentic systems that are not just impressive in demos, but stable, auditable, and trustworthy at scale.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.