Cloud infrastructure, built and documented.

TrailForge

Live storefront behind Cloudflare, on my own infrastructure

• Runs on my own DigitalOcean droplet behind Cloudflare with Full (strict) SSL; the origin answers on a separate direct hostname for comparison.
• Cache rules tuned so catalog and asset pages serve cache HITs while private and session pages bypass cache. A session cookie bug that briefly let private pages cache was found, isolated, and fixed.
• Free Managed WAF ruleset plus custom blocks for admin paths and SQL injection / XSS query patterns, and a login rate limit of 5 attempts per 10 seconds per IP, verified end to end from a browser.

• Cloudflare
• WAF
• Cache rules
• Rate limiting
• Nginx

Visit the live labRead the write-up

AWS ECS Fargate Deployment

Container deployment, health checks and task replacement

• Deployed an ECR-hosted container behind an Application Load Balancer with two running ECS Fargate tasks.
• Verified target health and task replacement, then practiced versioned updates using task revisions and a replacement service, comparing a mutable v2 rollout against an immutable v3 deployment.
• Documented deployment steps, architecture, and troubleshooting in an academic cloud lab.

• ECS Fargate
• Docker
• ALB
• ECR

View on GitHub

CallScribe + Prompter

Two apps, one folder: a local speech-to-text workflow

• CallScribe transcribes audio locally with Whisper models through faster-whisper, so no audio leaves the machine.
• Prompter takes the transcript and structures it into sections for live use, with an optional AI layer that can be switched off for a clean comparison.
• Evaluated both ways, AI off and AI on, and measured my own-voice transcription at about 19x real time with a 13.2 percent word error rate.

• Python
• Whisper
• Local AI
• Speech to text

Read the write-upPrompter on GitHub

Two AI Agents, One Notebook

Claude Code and Codex reviewing each other through one shared file

• An experiment in agent collaboration: two AI coding agents from two companies reviewing and correcting each other through a single shared notebook and one command line, with me as the only decision maker.
• Documents a real MCP tool failure and its actual cause, plus the checks that caught mistakes both agents made.

• AI agents
• MCP
• Workflows

Read the write-up

Service Selector

Solutions engineering workbench

• Independently built an AI-assisted workbench that connects discovery inputs, pain points, and risks to recommendations across a 77-product catalog.
• Added review gates for customer facts and assumptions, supporting statements, and product-plan compatibility checks.
• Created architecture views, configuration guidance, validation steps, downloadable reports, and editable presentations with speaker notes.

• Guided discovery
• Solution recommendations
• Architecture views
• Demo runbooks

Private project · capability overview

AWS CI/CD Pipeline

CodeBuild to S3 to CodeDeploy to EC2

• Configured CodeBuild to build and package a Maven application into S3 artifacts and CodeDeploy to deploy to EC2/Tomcat.
• Used IAM roles and documented SSH authentication, deployment paths, and troubleshooting.
• Included buildspec.yml, appspec.yml, application source, and deployment evidence in the repository.

• CodeBuild
• CodeDeploy
• EC2
• S3
• IAM
• Maven

View on GitHub

InsightFlow

Decision-ready answers from messy CSVs

• Built a CSV analytics app with fault-tolerant ingestion, schema validation, configurable cleaning, and statistical analysis.
• Added chart recommendations, data-quality flags, and exportable decision briefs with an optional AI narrative layer.

• CSV analytics
• Data validation
• Interactive charts
• Decision briefs

View on GitHub

SafeRoute Austin

Crash-risk analytics app

• Built a machine-learning app using 225K+ Austin crash records to score intersection risk and explain key safety factors.
• Created geospatial features, hotspot maps, batch CSV scoring, and map-based lookup for crash-risk analysis.

• Geospatial features
• Risk scoring
• Interactive maps

View on GitHub

AWS Incident Response Runbook

Desktop app for cloud ops

• Built a desktop app with guided incident-response checklists, metadata, progress tracking, and exports.
• Included AWS reference commands to support structured troubleshooting and documentation.

• TypeScript
• React
• Electron
• AWS

View on GitHub

Mattermost on AWS

Two-tier cloud architecture

• Deployed Mattermost and a self-managed MySQL database on separate EC2 instances in public and private subnets.
• Configured a custom VPC, route tables, Internet/NAT gateways, security groups, and SSH access.
• Documented the two-tier architecture, deployment process, and verification in a cloud course project.

• EC2
• VPC
• NAT Gateway
• MySQL
• Bash

View on GitHub

Neon Ark: Backend and CLI

Backend API and capstone system

• Built a Spring Boot backend with PostgreSQL, Flyway schema migrations, and a Java CLI client.
• Used layered components and input validation for a capstone creature-management system.

• Java
• Spring Boot
• PostgreSQL
• Flyway

View on GitHub

OS Teaching Tools

Interactive OS algorithm simulators

• Built five browser-based operating-system simulators with AI-assisted development and used them during ACC tutoring.
• Explained scheduling, resource allocation, and memory behavior through interactive visualizations and edge-case examples.

• Interactive simulations
• AI-assisted development
• Technical tutoring

View on GitHub

TrailForge: cache, attack, measure

How the TrailForge storefront is protected at the edge, and the session cookie bug that briefly cached private pages.

Read the post

Two Apps, One Folder

CallScribe and Prompter: a local speech-to-text workflow, tested with the AI layer off and on, and scored on my own voice.

Read the post

Two AI Agents, One Notebook

How I made Claude Code and Codex talk to each other: two AI agents, one shared notebook file and one command line, and a human who stays the only decision maker while they review and correct each other.

Read the post

All build notes

Every write-up, newest first, at blog.nasimlab.com. The pattern is the same everywhere: build it, measure it, write down what actually happened.

Visit the blog