Enkhbat E.

Multi-agent AI systems and data pipelines that survive production traffic. I build LLM platforms end to end: agent orchestration, RAG, FastAPI backends, and BigQuery and Spark pipelines on Google Cloud and AWS. Over 15 years I've built systems processing 1B+ records. 17 agents live at 98%+ reliability, sub-500ms. Expert-Vetted, one senior engineer, full system ownership. WHY ONE ENGINEER INSTEAD OF 3-4 SPECIALISTS One person is accountable for AI architecture, backend, frontend, LLM integration and cloud infrastructure. No handoffs, no integration gaps, no coordination overhead. WHAT I DO • AI and LLM: multi-agent orchestration (Google ADK, LangChain, LangGraph), model routing (Claude, Anthropic API, Gemini, OpenAI), RAG with vector search (Qdrant, Pinecone, pgvector), prompt engineering, AI automation, chatbots • Data engineering: BigQuery (load jobs, schema design, SQL, Terraform-managed datasets), ETL, data-quality scoring, Apache Spark, Kafka, PostgreSQL, MongoDB, data warehouse architecture • Backend: Python, FastAPI, Node.js, REST APIs, WebSockets, microservices, Stripe • Frontend: React, Next.js, Angular, TypeScript, SaaS dashboards • Cloud and DevOps: Google Cloud (Cloud Run, Vertex AI, Pub/Sub, BigQuery, Firestore), AWS (Lambda, ECS, EC2, S3), Terraform, Docker, Kubernetes, CI/CD RECENT WORK • Healthcare AI on Google Cloud: 17+ specialized agents on Google ADK and Vertex AI, with 98%+ reliability and sub-500ms P95 latency • Hospitality SaaS: multi-tenant platform for beach and resort operators, with a visual layout builder, a composable pricing engine and Stripe payments (Angular, FastAPI, PostgreSQL, Auth0) • Large-scale data pipeline: 1B+ eCommerce reviews across 87M products at 100M reviews/day on Apache Spark and AWS • BigQuery pipelines for AI Campus, a university recommendation platform: Python loaders from scraped JSON into BigQuery, then a quality-scored ETL into Firestore • BigQuery analytics at Kahani Health: GA4 pipeline with 18 event types behind executive dashboards • AI orchestration: 22 agents with QA loops, multi-model routing and runtime sandboxing • Algorithmic trading: 25+ automated systems handling 10,000+ requests per minute HOW I WORK Discovery and architecture review, clear milestones, documented sprints, and client sign-off on each delivery. Experienced with NDA and IP agreements, compliance documentation and background checks. Available for long-term work and retainers. Timezone UTC+8, with flexible overlap for US and EU business hours.