---

Triforge

Multi-tool IaC pipeline deploying an English to Spanish translation AI inference endpoint on AWS using Pulumi, Terraform, and OpenTofu across isolated infrastructure layers.

Architecture

Pulumi (Python)     → SageMaker endpoint + IAM + Secrets Manager
Terraform (HCL)     → EC2 instance + Security Groups
OpenTofu (HCL)      → S3 remote state backend
Flask (Python)      → Proxy server bridging UI ↔ SageMaker

State backend: S3 (remote-state-h1)
Config passing: AWS Secrets Manager → Flask → UI
Model: Helsinki-NLP/opus-mt-en-es via HuggingFace inference container

Why 3 Tools?

Each tool owns a layer that suits its philosophy:

• Pulumi — Python flexibility for SageMaker's complex conditional resources
• Terraform — battle-tested HCL for standard compute provisioning
• OpenTofu — FOSS Terraform fork managing shared state infrastructure (i honestly just wanted to try it out as it looked kinda cool)

Structure

triforge/
├── sagemkaer/        # Pulumi — SageMaker + Secrets
├── compute/          # Terraform — EC2
├── storage/          # OpenTofu — S3
└── ui/
    ├── proxy/        # Flask proxy app
    └── public/       # HTML/CSS/JS frontend

Prerequisites

• AWS CLI configured
• Pulumi CLI, Terraform, OpenTofu installed
• Python 3.10+, pip

Deploy

# 1. Storage (OpenTofu)
cd storage && tofu init && tofu apply

# 2. SageMaker (Pulumi)
cd sagemkaer && pulumi up

# 3. Compute (Terraform)
cd compute && tf init && tf apply

# 4. UI runs via CI/CD on push to main

Links

Blog @ Triforge

Github @ Triforge-Code

Live @ Triforge-chat

---