I am an AI Lead & Architect at Oracle who designs and delivers enterprise-grade Generative AI systems — end to end, from architecture to production.
My work spans the full AI engineering spectrum: RAG pipelines, multi-agent orchestration, LLM-powered automation, video and document intelligence, and AI platforms built on top of enterprise data — deployed on OCI (Oracle Cloud Infrastructure) and integrated with the Oracle technology ecosystem.
- 🏗️ Architecting enterprise RAG systems — hybrid search, reranking, multi-turn chat, source citations, knowledge copilots
- 🤖 Designing Agentic AI frameworks — multi-agent orchestration, planning, tool use, memory, and human-in-the-loop
- 🎬 Building AI Video & Document Intelligence — Gemini-powered video analysis, transcription, content moderation, meeting intelligence
- 🔐 Delivering domain AI platforms — fraud detection, medical AI assistants, personalized learning OS, financial intelligence
- ⚙️ Engineering AI-powered automation — workflow orchestration, document processing, intelligent classification pipelines
- ☁️ Deploying AI workloads on OCI — OCI Generative AI, OCI Data Science, Oracle Autonomous DB, OCI Object Storage
"I don't just build AI features — I architect AI systems that are observable, scalable, and trusted in production."
╔══════════════════════════════════════════════════════════════════════════╗
║ ENTERPRISE AI PLATFORM — FULL ARCHITECTURE VIEW ║
╠══════════════════════════════════════════════════════════════════════════╣
║ ║
║ INPUT LAYER AI CORE OUTPUT LAYER ║
║ ┌────────────┐ ┌─────────────────┐ ┌──────────────┐ ║
║ │ Documents │──────────▶│ RAG ENGINE │───────▶│ AI Chat / │ ║
║ │ PDFs, DOCX │ │ Hybrid Search │ │ Copilot UI │ ║
║ │ Web, APIs │ │ Reranking │ └──────────────┘ ║
║ └────────────┘ │ OCI Search / │ ║
║ │ pgvector/Pine. │ ┌──────────────┐ ║
║ ┌────────────┐ └────────┬────────┘ │ AI Agents │ ║
║ │ Videos │──▶ Gemini 2.0 ──▶│ LLM LAYER │ Research │ ║
║ │ Meetings │ Flash / │ Claude/GPT/ │ Writing │ ║
║ │ Audio/STT │ Whisper │ OCI GenAI │ Support │ ║
║ └────────────┘ │ Multi-model │ Onboarding │ ║
║ ┌────────┴────────┐ └──────────────┘ ║
║ ┌────────────┐ │ AGENT ENGINE │ ║
║ │ Oracle DB │──────────▶│ Plan→Act→Obs │ ┌──────────────┐ ║
║ │ Auto. DWH │ text2sql │ Tool Use │───────▶│ Workflow │ ║
║ │ Snowflake │ │ Memory (ST/LT) │ │ Automation │ ║
║ │ BigQuery │ │ Human-in-Loop │ │ Pipelines │ ║
║ └────────────┘ └─────────────────┘ └──────────────┘ ║
║ ║
║ INFRA: OCI · FastAPI · Oracle DB · Redis · Docker · K8s · CI/CD ║
╚══════════════════════════════════════════════════════════════════════════╝
| Project | What It Does | Key AI Stack |
|---|---|---|
| 🤖 Enterprise RAG Knowledge Copilot | Company-wide AI brain — chat, voice, meetings, 20+ connectors, hybrid search, multi-agent | Claude · FastAPI · OCI · pgvector · Next.js |
| 🎓 AI Personalized Learning OS | Adaptive learning — AI curriculum generation, skill gap analysis, personalized paths | LLM · Python · LangChain · RAG |
| 🏥 AI Medical Assistant & Patient Monitor | Clinical AI — symptom analysis, patient Q&A, real-time monitoring, medical document RAG | Claude · Python · FastAPI · RAG |
| Project | What It Does | Key AI Stack |
|---|---|---|
| 💳 AI Financial Intelligence & Fraud Detection | Real-time fraud scoring, anomaly detection, AI-driven financial insights | Python · LLM · ML · OCI Data Science |
| 🎬 AI Video Intelligence & Content Moderation | Video analysis, transcript generation, content classification, scene understanding | Gemini 2.0 Flash · Whisper · Python |
| 🧑💼 AI Interviewer & Skill Assessment Platform | Automated AI interviews, competency evaluation, skill scoring, hiring intelligence | LLM · TypeScript · Next.js · RAG |
| Project | What It Does | Key AI Stack |
|---|---|---|
| 🤝 Agentic AI Workflow Orchestrator | Multi-agent framework — Research, Writing, Support, Onboarding, Data Analyst agents | LangGraph · CrewAI · Python · Redis |
| 📊 AI-Powered Data Intelligence Platform | Text-to-SQL on Oracle ADW / DWH, AI insight generation, data quality agents | Claude · Oracle ADW · dbt · Python |
expertise = {
"RAG & Knowledge Systems": [
"Hybrid Search (Dense Vector + BM25 Full-text)",
"Cross-encoder Reranking", "Semantic Chunking",
"Multi-turn Conversational RAG", "Source Citation & Grounding",
"Oracle AI Vector Search · OCI OpenSearch · pgvector · Pinecone",
"20+ Data Connectors (Drive, Slack, Notion, Jira, Confluence)"
],
"Agentic AI & Orchestration": [
"Multi-agent Systems (ReAct, Plan-and-Execute)",
"LangGraph · CrewAI · AutoGen",
"Tool Use (Web Search, Code Exec, DB Query, API Calls)",
"Short-term & Long-term Memory", "Human-in-the-loop",
"Research · Writing · Support · Onboarding · Data Analyst Agents"
],
"LLM Engineering": [
"OCI Generative AI Service (Cohere, Llama on OCI)",
"Multi-model Routing (Claude · GPT-4 · Gemini · OCI GenAI)",
"Prompt Engineering & Chain-of-Thought",
"Streaming SSE Responses", "LLM Evaluation & Hallucination Detection",
"Cost Optimisation · Guardrails & Safety Layers"
],
"Video & Multimodal AI": [
"Gemini 2.0 Flash Video Analysis (visual + audio + slides)",
"OpenAI Whisper Transcription", "Meeting Intelligence",
"AI Content Moderation", "Scene Understanding",
"Video Q&A and Chapter Generation"
],
"OCI & Oracle Stack": [
"OCI Generative AI Service", "OCI Data Science (Model Deployment)",
"Oracle AI Vector Search (23ai)", "Oracle Autonomous Data Warehouse",
"OCI Object Storage", "OCI OpenSearch",
"Oracle Database 23ai", "OCI Container Instances / OKE"
],
"AI System Architecture": [
"End-to-end AI Platform Design (PoC → Production)",
"FastAPI Async Backends · SSE Streaming",
"Multi-tenant Architecture · OAuth (Google, Microsoft)",
"Redis Caching · Connection Pooling · DB Indexing",
"OKE · ArgoCD GitOps · Terraform · Prometheus · OpenTelemetry"
],
"Data & DWH (AI-enabling layer)": [
"Text-to-SQL on Oracle ADW / Snowflake / BigQuery",
"AI Data Quality Agents", "Medallion Architecture",
"dbt · SQL · Airflow", "Automated Insight Generation on DWH"
],
}1. Architecture outlives models. The retrieval strategy, agent design, and data layer matter more than which model you pick today.
2. Observability is not optional. If you can't trace your RAG pipeline or agent chain, you cannot debug or improve it.
3. Agents need guardrails. Autonomous systems require human-in-the-loop checkpoints, cost controls, and fallback paths — always.
4. Multimodal is the present, not the future. Text, voice, video, and structured data must all flow into one unified knowledge layer.
5. PoC ≠ Production. Latency, cost, reliability, and security are first-class citizens from day one — not afterthoughts.
- 🧪 RAG Evaluation Framework — automated faithfulness scoring, hallucination detection, answer relevance benchmarking
- 🔗 Multi-model AI Router — intelligent routing across Claude / GPT-4o / OCI GenAI based on task type and cost
- 🎬 Video Intelligence Pipeline — Gemini 2.0 Flash powered enterprise video RAG with chapter markers and Q&A
- 🏛️ Oracle 23ai Vector Search RAG — native vector embeddings inside Oracle Database 23ai for enterprise knowledge retrieval
Open to discussing AI architecture, enterprise GenAI strategy, and principal / lead / architect roles where I can drive end-to-end AI system design.