Abhishek Kumar Gupta

Senior Generative AI Engineer with 5+ years of software engineering experience and 2.5+ years building production-grade Agentic AI and RAG systems in regulated BFSI environments. Hands-on experience designing and deploying multi-agent systems using LangGraph, LangChain, CrewAI, and AutoGen, with tool/function-calling integrations to backend and data systems via MCP (Model Context Protocol) and REST APIs. Strong track record across the full agentic AI lifecycle — architecture, retrieval, orchestration, fine-tuning (LoRA/QLoRA/PEFT), evaluation (RAGAS, DeepEval, LangFuse), and cloud-native deployment on Azure and AWS — delivering measurable gains in accuracy, latency, and cost across enterprise GenAI products.

Author of a peer-reviewed workshop submission on failure modes in production multi-agent LLM systems.

Tech Lead — Generative AI & Agentic AI Engineering at Cognizant (2022-12 – Present)

Leading design and delivery of Agentic AI and RAG systems for a regulated financial services client

• Lead the design and development of Agentic AI and RAG systems for BFSI workflow automation, owning architecture decisions from ideation through production deployment with minimal supervision.
• Designed and deployed a multi-agent, LLM-driven financial volatility prediction system using Python, Azure OpenAI, and agentic orchestration patterns, with real-time data ingestion via PySpark for automated risk scoring.
• Engineered an enterprise RAG pipeline over 100K+ records using hybrid retrieval, reranking, and semantic caching, improving response accuracy by 45% over legacy baselines and cutting LLM token cost by 35%.
• Fine-tuned domain-specific LLMs using LoRA, QLoRA, and PEFT, applying knowledge distillation to reduce inference compute cost by 50% while maintaining over 95% task accuracy.
• Improved real-time agentic service performance through retrieval optimization, semantic caching, and microservices tuning, achieving 60% faster retrieval.
• Built evaluation and observability pipelines using RAGAS, DeepEval, and LangFuse to continuously monitor agent/LLM latency, output quality, and cost across production services.
• Modernized data workflows for AI-ready pipelines, migrating legacy MQ systems to Kafka to support real-time, agent-driven analytics.
• Partnered with DevOps on CI/CD enablement for agentic AI services using Terraform and Jenkins, strengthening infrastructure-as-code practices and deployment reliability.
• Contributed to large-scale cloud-native architecture across AWS and Azure, focused on scalability, microservices integration, and real-time multi-agent system performance.

Middleware Engineer — Messaging & Integration (IBM MQ) at Cognizant (2020-10 – 2022-11)

Configured, monitored, and supported enterprise messaging pipelines for critical business systems

• Configured and administered IBM MQ queue managers, queues, and channels supporting enterprise messaging pipelines across critical business systems.
• Debugged failed, stuck, and misrouted messages across MQ-based integration pipelines, tracing issues end-to-end across producers, queues, and consuming applications to restore reliable data flow.
• Monitored MQ infrastructure health and channel status, proactively identifying bottlenecks and message backlogs before they impacted downstream processing.
• Led incident response and root-cause analysis for middleware failures, coordinating across application and infrastructure teams to restore service and prevent recurrence.
• Worked within SDLC delivery practices, supporting cloud based integration systems, troubleshooting, and secure coding.

B.Tech in Electronics & Communication Engineering (2016 – 2020)