Agentic Systems

Description Developed a mobile application tailored for TOEFL education institutions. Designed a scalable AWS Cloud infrastructure capable of supporting 500–600K users, aligned with the client’s budget constraints. Managed end-to-end CI/CD pipelines and organized deployment workflows. Independently designed and built the entire cloud architecture from scratch using core AWS services including SQS, S3, ECS, VPC, Secrets Manager, ALB, Route 53, CloudFront, and CloudWatch. Led and coordinated a 3-member engineering team throughout the project lifecycle. Conducted direct client meetings and performed R&D on integrating TOEFL-specific educational content into the software product. Developed AWS Lambda-based applications to deploy and run AI models within the system. AI & Agentic Systems (Current Role) AI Orchestration & Agentic Systems: Connecting Claude via APIs to internal databases, compliance guardrails, and custom workflows using agent orchestration platforms. Custom Integration: Working directly within a client’s legacy codebase to design custom Retrieve-Augmented Generation (RAG) pipelines and sub-agent hierarchies. Field Intelligence: Providing high-level technical support, capturing customer use cases, and feeding feedback directly to Anthropic’s core product teams. Frontend & Backend Development Built both frontend and backend features using TypeScript. Improved software quality by writing comprehensive backend tests using Vitest. Developed a CMS panel ensuring it met technical content and operational requirements. Key Skills Cloud Infrastructure: AWS (SQS, S3, ECS, VPC, Secrets Manager, ALB, Route 53, CloudFront, CloudWatch), Docker, Kubernetes AI & Agentic Workflows: Claude API integration, RAG pipelines, multi-agent orchestration frameworks CI/CD & DevOps: End-to-end pipeline management, deployment workflows, system observability Production Engineering: TypeScript (frontend & backend), Vitest testing Team Leadership: Led 3-member engineering team, client-facing technical communication

Overview I design and deploy practical AI solutions that automate real business workflows - not generic chatbots, but purpose-built systems connecting your internal data, compliance rules, and operational processes with large language models. From hooking Claude or GPT into your application to building multi-agent orchestration pipelines, I handle the integration layer so AI works naturally within your existing systems. What I Can Do For You LLM and API Integrations Model integration: Connect Claude, GPT, Gemini, or local open-source models to your application via REST APIs or SDKs Context management: Conversation history, session management, and context window optimization Structured output: Parse LLM responses into typed, validated data your application can consume reliably Fallback chains: Graceful degradation across multiple model providers for high availability Retrieval-Augmented Generation (RAG) Knowledge base QA: Question-answering over your company documentation, policies, product catalogs, and technical manuals Vector search: Embedding pipelines with vector databases for semantic search over large document collections Real-world example: TOFY Go! - custom RAG pipelines for TOEFL-specific educational content delivery at Bee2 AI Agentic Systems Multi-agent orchestration: Hierarchies of specialized AI agents - one retrieves data, another analyzes, another formats output Sub-agent delegation: Complex tasks broken into parallel sub-tasks handled by independent AI agents Human-in-the-loop: Critical decisions routed to human approvers before execution Business Process Automation Document processing: Auto-extract, classify, and route invoices, contracts, and forms Reporting automation: AI-generated daily/weekly reports from structured and unstructured data Workflow triggers: Event-driven automation (new order, AI checks inventory, generates response) MCP (Model Context Protocol) Servers Custom MCP servers: Expose your internal tools, databases, and APIs as MCP tools for AI assistants Development tools: Custom MCP servers for code analysis, deployment automation, and DevOps Technologies LLMs: Claude API, OpenAI API, Gemini, open-source models (Llama, Mistral, Qwen) Frameworks: LangChain, custom agent orchestration Data: RAG pipelines, vector databases (Pinecone, Qdrant), structured SQL Protocol: Model Context Protocol (MCP) Integration: REST APIs, webhooks, message queues Deployment: Docker, AWS Lambda, serverless Languages: TypeScript, Python