Aditya Kumar

Cloud Accelerator Team, JPMorgan Chase

Helix — Multi-Agent Intelligence Platform

• Architected and developed Helix, a multi-agent intelligence platform with 17 specialized AI agents orchestrated through a fault-tolerant concurrent pipeline, querying 7 MCP servers (SEAL, V12, Splunk, GAP, Stratus, JIRA, Confluence) to synthesize cross-platform infrastructure data in real time
• Designed a phased pipeline architecture with router-based query classification, concurrent data gathering via asyncio, and sequential post-processing — reducing follow-up query latency from ~60s to ~3-5s using FAISS-based session caching with offline feature-hashing embeddings
• Built 7 RFC 9728-compliant OAuth 2.0 MCP servers (REST + GraphQL) for SEAL, V12, Splunk, GAP, Stratus, Camp Insights, and Dynatrace APIs with automatic IDA token management, connection pooling, and Docker Compose orchestration
• Implemented cross-MCP data chaining for agents to extract join keys (SEAL IDs, app GUIDs, hostnames) to correlate application metadata, infrastructure, logs, deployments, and Kubernetes hygiene across disparate data sources, enhancing data accessibility for analysis
• Developed real-time Gradio dashboard supporting 12 dynamic component types (metric cards, gauge charts, heatmaps, service maps, timelines) with streaming agent event visualization and per-MCP drill-down analytics, improving user insights into system performance
• Integrated Azure OpenAI (GPT-4o) with CRT-based Azure AD authentication and Arize Phoenix OpenTelemetry for agent observability and tracing

FarmFix Remediation — VS Code Extension

• Developed a VS Code extension integrating SiC secrets detection, Snyk SCA/SAST scanning, EAC ModulePack upgrades, and GAP Buildpack management into a unified sidebar with tree views, findings navigation, and a webview report dashboard
• Implemented Copilot Chat agent integration dispatching scan and remediation workflows to AI-powered agents for automated security finding resolution
• Managed internal distribution workflow including SEAL registration, PAR submission, and JPMC internal VS Code Marketplace onboarding

Safe Deployment Strategy — Blue-Green Patterns

• Designed and implemented 8 safe deployment patterns for Spring Boot microservices across EKS, GAP, and ECS platforms using EAC and Spinnaker, targeting zero-downtime releases
• Developed EKS Blue-Green Infrastructure pattern enabling zero-downtime EKS cluster upgrades, and EKS Application Blue-Green pattern with Post-Deployment Validation Testing (TrueCD) and automated rollback
• Led POC with Spinnaker team for functional testing stage integration, enabling pre-traffic-switch test execution to validate infrastructure and environment readiness before live cutover
• Configured GAP-GKS multipool deployment pattern with GAIA Kafka Service integration, including Sophia CIDA/ESaaS onboarding, topic creation, producer/consumer setup, and isolated consumer group validation
• Implemented ECS-GKS blue-green connectivity pattern requiring C2C AWS accounts, FID OIDC configuration, and multi-firewall request coordination
• Authored comprehensive Confluence documentation for 8 patterns, providing a reference for application teams to adopt safe deployment strategies, streamlining knowledge transfer and implementation processes

RGEN — Automated Runbook Generation

• Built an agentic runbook generation and review system using CDAOSMART SDK with MCP integrations to Confluence, JIRA, SEAL, ServiceNow, Gaia Portal, and CloudVision for automated technical documentation
• Implemented orchestrator-driven feedback loops with a RAG agent backed by a knowledge vector database for enhanced information retrieval and continuous runbook improvement
• Mentored 3 interns on agentic AI development, guiding them through CDAOSMART platform architecture and code implementation across v0.0.7 and v1.0.1

Ignite — Agentic KAG & FinOps

• Built a Knowledge-Augmented Generation (KAG) demo using CDAOSMART, Neo4j, and graph transformers to ingest Confluence data into a graph database for AI-powered retrieval
• Developed a FinOps dashboard automating CFM tracker data extraction and visualization for cloud cost management

Designed on-prem-cloud hybrid cluster with three managed K8s master nodes on AWS EKS and two on-premise nodes.
Migrated critical applications such as NetBox, ClickHouse, Jenkins, AWX, and FreeIPA to new infrastructure.
Automated application deployment using GitOps principles with ArgoCD for improved efficiency.
Developed tailored manifest and Helm charts specific to use cases and custom CRDs for the cluster.
Implemented dynamic PVC provisioning via Kubernetes NFS provisioner by directly invoking storage class.
Managed on-prem cluster endpoint with Keepalived and HAProxy for high availability.
Established disaster recovery strategies supporting n-3 (read-write) and n-4 (read-only) configurations.
Executed migration of over 600 repositories and 150 projects from Bitbucket Server to Bitbucket Cloud.

Streamlined deployment strategies through automation using Slack bots and Terraform for AWS/Azure.
Facilitated efficient deployments in both development and production phases.
Deployed containerization solutions leveraging ECS and EKS within AWS environment.
Managed Infrastructure as Code (IaC) using Terraform and shell scripts backed by S3 storage.
Created Slack bots to enable swift scaling of development infrastructure, optimizing cost efficiency.
Captured key learnings on Log4j RCE vulnerability through documentation on GitBook.
Constructed infrastructure for Cactus editing tool using Terraform, managing deployments via Helm charts.
Enhanced security with WAF/ACM integrations into CloudFront and leveraged NLB for traffic routing.

• Managed cloud infrastructure using AWS and Azure platforms.
• Automated deployment pipelines using Jenkins and GitLab CI/CD tools, enhancing release consistency and speed.
• Collaborated with development teams to integrate processes and ensure continuous delivery, facilitating smoother release cycles.

Designed architecture for Morningstar Asset Management Dashboard, facilitating predictive fund analytics.
Explored AWS services to optimize pipeline flow, addressing specific project requirements.
Automated processes using various Lambda functions for decision-making in data pipelines.
Initiated project from scratch, achieving production approval with minimal modifications.
Executed ETL processes using Python and Anaconda on Spot EC2 instances, reducing operational costs.
Scheduled jobs via Jenkins for Sonar analysis and unit testing prior to deployment.
Implemented DynamoDB to track stage statuses and manage cooldown periods for sensitive data processing.
Integrated SNS notifications for step completion and escalated monitoring through VictorOps.

• Collaborated with development teams to streamline CI/CD processes, enhancing deployment efficiency.
• Managed cloud infrastructure on AWS for enhanced scalability and reliability.
• Designed architecture for Morningstar Asset Management Dashboard, facilitating predictive fund analytics.Explored AWS services to optimize pipeline flow, addressing specific project requirements.Automated processes using various Lambda functions for decision-making in data pipelines.Initiated project from scratch, achieving production approval with minimal modifications.Executed ETL processes using Python and Anaconda on Spot EC2 instances, reducing operational costs.Scheduled jobs via Jenkins for Sonar analysis and unit testing prior to deployment.Implemented DynamoDB to track stage statuses and manage cooldown periods for sensitive data processing.Integrated SNS notifications for step completion and escalated monitoring through VictorOps.