Babatunde Adejumo

Experience

1. Agentic Ai Architect Senior Mlops Engineer

   Cognizant

   Sep 2021 — Present · US

   Designed and deployed enterprise-grade multi-agent systems using CrewAI, LangGraph, LangChain, and custom orchestration frameworks to automate complex workflows, documentation intelligence, compliance tasks, and decision-making in regulated environments.Built adaptive LLM-driven systems with planning, memory, tool execution, contextual routing, and hierarchical reasoning for high-stakes automation across finance, ERP, CRM, and internal operations.Architected advanced Retrieval-Augmented Generation (RAG) pipelines using Azure AI Search, Weaviate, and FAISSEnabled HyDE, adaptive prompting, re-ranking, and custom embeddings to boost structured generation quality and reasoning accuracy.Built scalable MLOps pipelines with MLflow, SageMaker, Kubeflow, Docker, and GitLab CI/CD for continuous model deployment, evaluation, versioning, and rollback.Implemented model monitoring with CloudWatch, Langfuse, LangSmith, SNS, and EventBridge—adding guardrails for drift, latency, quality, and hallucination tracking.Fine-tuned and optimized LLMs (GPT-3.5, Llama 2, Mixtral) using LoRA/QLoRA, Quanto, quantization, and proprietary domain datasets.Delivered a guardrail-aware architecture ensuring compliance and reliability for sensitive workloads.Built an LLM observability suite using Langfuse + RAGAS + custom metrics with dashboards for latency, cost, semantic correctness, relevance, groundedness, and agent reasoning traceability.Architected AI-ready cloud infrastructure using AWS EKS, Lambda, DynamoDB, SQS/SNS, API Gateway, SageMaker, and S3 for scalable inference, training, embeddings generation, and multi-agent workload orchestration.Developed event-driven image and data processing pipelines using AWS Lambda, S3 triggers, Glue ETL, Crawlers, Athena, and Redshift.Reduced enterprise operational workload by 40% using autonomous multi-agent systems.Accelerated AI adoption by creating reusable agent templates and orchestration blueprints used across practice teams.