← Back to About

Cloud Architect Intern

Oracle

Jun 2024 – Sep 2024 • Austin, Texas

Developed a production-grade Generative AI + RAG chatbot for Oracle Identity Governance, automating policy workflows and reducing access resolution time from 24h to 2h.

Key Accomplishments

  • ▸ Built RAG chatbot integrated with Oracle Identity Governance (OIG), indexing 1,000+ policy docs for contextual search; achieved 95% accuracy and <2s response time.
  • ▸ Automated access-request workflows via OIM REST APIs, cutting manual resolution from 24h to 2h and handling 50+ daily requests.
  • ▸ Optimized retrieval pipeline with parallel search and caching, improving latency by 4× while maintaining scalability for 100+ concurrent users.
  • ▸ Engineered OCI-based monitoring dashboards and alerting for OIM services, improving system reliability and proactive issue detection.

Skills & Technologies

  • Python
  • RAG
  • LLM
  • OCI
  • Identity Governance
  • FastAPI
  • LangChain

What I did

Designed and deployed a Retrieval-Augmented Generation (RAG) chatbot that integrates directly with Oracle Identity Governance (OIG), enabling real-time identity self-service through natural language interaction.

Why it mattered

  • Reduced access-request turnaround from 24 hours to 2 hours.
  • Automated manual policy lookups (previously 10+ minutes) to instantaneous (<2s) results.
  • Enabled self-service for 50+ access requests/day, reducing operational load on support teams.

How it worked

  • RAG pipeline: Indexed 1,000+ OIG documents into a vector store using embeddings (Cohere Command R+), enabling precise semantic retrieval.
  • Chat orchestration: Implemented multi-turn conversation flow for complex identity and access queries.
  • Workflow automation: Integrated with OIM REST APIs for access requests, approval routing, and policy updates.
  • Latency optimization: Used asynchronous parallel retrieval, in-memory caching, and streaming responses to cut latency by 75%.
  • Monitoring: Deployed OCI-based observability dashboards with custom metrics and alerts for OIM performance tracking.

Tech

  • Backend: Python, FastAPI, LangChain
  • Vector DB: Oracle Autonomous DB 23ai, FAISS
  • LLM: Cohere Command R+ via Oracle Generative AI
  • Infrastructure: OCI Kubernetes, OCI Functions, OCI Monitoring

My role

  • Architected the full RAG pipeline, including data ingestion, embedding generation, retrieval optimization, and context assembly.
  • Built the chatbot engine with multi-step reasoning and API-based automation for Oracle Identity Manager (OIM).
  • Developed OCI monitoring and alerting for OIM services, integrating performance metrics into custom dashboards.