~/ renganayaki-iswarya / projects

Projects

AI agents, RAG-ready backends, and the cloud infrastructure that ships them. Each project below includes the repo, the stack, and the structure at a glance.

#01Solo — design, prompts, backend, deploy

Bilingual Tour Operations Agent

Structured data extraction from unstructured travel enquiries

An AI agent that reads travel enquiries in Chinese or English, extracts structured details (dates, budget, group size) via prompt engineering, and flags incomplete enquiries for human review.

  • Bilingual intake (ZH / EN) with structured output schema
  • Human-in-the-loop flagging for incomplete enquiries
  • Roadmap: Qdrant · RAG · LangGraph · JWT · Admin Dashboard
FastAPIGeminiPrompt EngineeringPydanticPostgreSQLDocker
View repository
structure
bilingual-tour-ops-agent/
├── app/
│   ├── main.py            # FastAPI entrypoint
│   ├── agent/
│   │   ├── prompts.py     # bilingual system prompts
│   │   └── extractor.py   # structured extraction
│   ├── schemas/           # Pydantic models
│   └── db/                # PostgreSQL layer
├── Dockerfile
└── tests/
#02Solo — agent design, integrations, deployment

Enterprise Customer Service Agent — Yellamma Bot

Configurable assistant for appointments & business workflows

A general-purpose, configurable customer service assistant reusable across businesses. Handles conversations, appointment scheduling, and business workflows via FastAPI, LLM APIs, and prompt engineering.

  • Configurable prompt profiles per business
  • Appointment scheduling workflow end-to-end
  • WhatsApp channel integration
FastAPILLM APIsWhatsApp APIPrompt EngineeringPython
View repository
structure
yellamma-bot/
├── bot/
│   ├── router.py          # FastAPI routes
│   ├── flows/
│   │   ├── booking.py     # appointment logic
│   │   └── faq.py
│   ├── prompts/           # per-tenant configs
│   └── channels/
│       └── whatsapp.py
└── deploy/
#03Solo — infra, IaC, CI/CD

Cloud-Native AI Deployment Platform

Containerized API on AWS EKS — two ways

Deployed a containerized API to AWS Kubernetes (EKS), built twice — with eksctl and with Terraform — with autoscaling, health checks, CI/CD, and documented cost & security trade-offs.

  • Two parallel provisioning paths (eksctl vs Terraform)
  • Horizontal Pod Autoscaler + liveness/readiness probes
  • Documented cost & IAM security posture
AWS EKSTerraformeksctlDockerKubernetesCI/CD
View repository
structure
FINTECH-CLOUD-EKS/
├── terraform/
│   ├── vpc.tf
│   ├── eks.tf
│   └── iam.tf
├── eksctl/
│   └── cluster.yaml
├── k8s/
│   ├── deployment.yaml
│   ├── service.yaml
│   └── hpa.yaml
└── .github/workflows/
#04Solo — full pipeline

DevOps Infrastructure Automation

Terraform → Ansible → Jenkins, push-to-deploy

End-to-end DevOps pipeline: Terraform provisions infrastructure, Ansible configures the server, Jenkins automates build-and-deploy on every push. Connects AI systems into existing company environments.

  • Immutable infra with Terraform state management
  • Idempotent server config via Ansible playbooks
  • Jenkins pipeline triggered on every push
TerraformAnsibleJenkinsAWSGitHub Actions
View repository
structure
devops-stack/
├── terraform/
│   └── main.tf            # EC2 · VPC · SG
├── ansible/
│   ├── playbook.yml
│   └── roles/
└── jenkins/
    └── Jenkinsfile        # build → test → deploy
→ roadmapin progress

Planned Enterprise AI Platform

Consolidating the projects above into one deployable enterprise stack.

enterprise-ai-platform
enterprise-ai-platform/
├── ai-knowledge-base/
│   ├── pdf-upload
│   └── rag-search
├── ai-agent/
│   └── multi-language-support
├── backend/
│   ├── fastapi
│   └── postgresql
├── infra/
│   ├── docker
│   ├── aws-deployment
│   └── ci-cd
├── security/
│   └── jwt-authentication
└── ops/
    ├── admin-dashboard
    ├── logging
    └── monitoring