swimstream.
Ask the AI Portfolio
Live Demos What I Build Work Stack AI Philosophy Ask the AI Portfolio
new Don't just read my portfolio. Query it.

AI-integrated software systems for complex business logic.

Swimstream builds production web applications, APIs, calculators, dashboards, and AI-enhanced workflows for businesses that need more than a brochure site.

Ask the AI Portfolio Explore Live System Demos
Fintech Calculators SaaS LMS Systems AI Knowledge Bases API Platforms PHP / MySQL Python / FastAPI PostgreSQL / pgvector
live workflow → software /api/workflow/plan
generated_plan.json · sample
# input: "loan officers re-calculate amortization by hand" modules: ["Calculator Engine", "Scenario Store", "REST API"] tables: ["loan", "scenario", "rate_table", "audit_log"] endpoints: POST /api/loans/amortize ai_assist: "explain scenario deltas" (human-reviewed) phases: 4 # model → api → ui → ai

mock Try your own above — output is generated client-side. Real planner connects to FastAPI.

Live AI Systems running client-side · backend-ready

The site is the proof. Use it.

Four working demos of the kind of AI-integrated software Swimstream builds. Every response is generated here in the browser — each one is a clean seam where a real FastAPI + Claude/OpenAI service plugs in.

demo_02

Workflow → Software Generator

Describe a manual process. Get modules, a data model, API endpoints, AI opportunities, and a build plan.

try:
your generated system blueprint appears here
demo_01

Ask My Portfolio

A RAG assistant grounded in Tim's work. Answers cite their source.

Ask me about Swimstream's work, stack, or how AI could fit your business. Answers are grounded in source docs.
index.mdservices.md
demo_03

API Response Builder

Structured responses with result codes, tokens, validation, and metadata.

copy
response.json
demo_04

AI Code Review Theater

Messy code → AI-assisted analysis → refactor → human approval. AI never merges alone.

UserRepository.php · submitted
function getUser($id) { // pulled straight from the request $q = "SELECT * FROM users WHERE id = " . $_GET['id']; return mysql_query($q); }
What Swimstream Builds

Real business systems — not brochure sites.

Ten things Swimstream turns into reliable, maintainable software. Pick the one that sounds like your problem.

Fintech calculator platforms

Financial logic as structured APIs — scenarios, validation, and business rules kept testable.

API systems

Versioned, validated REST APIs with predictable result codes and clean schemas.

SaaS LMS platforms

Database-driven learning systems: courses, roles, admin workflows, reporting.

Internal dashboards

Operational views that replace spreadsheets — filters, status, exports.

Knowledge-base chatbots

Source-grounded assistants that answer from your docs, with citations.

RAG / search systems

Embeddings + vector search to make specialized knowledge queryable.

Workflow automation

Manual processes become forms, APIs, and AI-assisted steps.

Legacy PHP/MySQL modernization

Stabilize, refactor, and extend production systems people depend on.

AI-assisted dev systems

Tooling and pipelines where AI accelerates work behind human review.

PostgreSQL / pgvector tools

Hybrid relational + vector search for fast, grounded retrieval.

Case Studies view template →

Problem → System → Result → Stack.

A consistent structure for every engagement: what was broken, what got built, what it changed, and what it runs on.

Fintech 01

Fintech Calculator Platform

ProblemLoan scenarios were re-computed by hand in spreadsheets — slow, inconsistent, impossible to audit.
SystemA calculator engine with scenario handling, structured API responses, server-side validation, and versioned business rules.
ResultConsistent figures, an auditable rule set, and an API the rest of the product can call.
StackFastAPIPostgreSQLREST
Read the build
SaaS · LMS 02

SaaS LMS System

ProblemAn aging learning platform was hard to maintain and slow to extend with new admin needs.
SystemRebuilt around a clean schema with role-based admin workflows, reporting, and long-term maintainability in mind.
ResultA platform that's safe to change — new features ship without fear of breaking the core.
StackPHPMySQLJavaScript
Read the build
AI · RAG 03

AI Knowledge Base / RAG Prototype

ProblemSpecialized knowledge lived in PDFs and email — slow to find and impossible to trust at a glance.
SystemFastAPI + PostgreSQL/pgvector with embeddings and source-grounded responses — every answer cites its chunk.
ResultKnowledge becomes queryable and verifiable, with a human-review path for sensitive content.
StackpgvectorembeddingsClaude
Read the build
Automation 04

Business Workflow Automation

ProblemA core process ran on manual steps, scattered data, and tribal knowledge — nothing was repeatable.
SystemTurned the process into dashboards, forms, and APIs, with AI-assisted steps where they measurably saved time.
ResultA repeatable system one person can run — with an audit trail instead of guesswork.
StackPHPPythonREST
Read the build
Technical Stack

Boring where it counts. Modern where it pays.

PHP/MySQL for systems people depend on, Python/FastAPI + PostgreSQL/pgvector for AI services. Chosen for reliability and maintainability — not novelty.

Backend & Data
PHP MySQL Python FastAPI PostgreSQL pgvector
Web & Interface
JavaScript jQuery HTML CSS / SCSS Tailwind
APIs & Retrieval
REST APIs JSON Embeddings RAG Vector search
AI Models
Claude OpenAI Local AI models
Infra & Ops
Git cPanel / WHM Linux

A typical AI request path. Retrieval is grounded in your data; the model drafts; a person stays in the loop on anything that matters.

AI Philosophy

AI where it earns its place.
Engineering everywhere else.

Swimstream uses AI to move faster and document better — never as a substitute for the judgment that keeps business-critical systems safe.

Validated Tested Human-reviewed
01
AI accelerates, judgment decides

AI speeds up development, but it does not replace engineering judgment. The hard calls stay with a person who understands the consequences.

02
Business-critical means reviewed

Anything that touches money, data, or decisions gets validation, testing, and human review before it ships.

03
AI improves accuracy, not just speed

Used well, AI raises workflow accuracy, documentation quality, and insight — not just raw output.

04
Privacy and production safety are non-negotiable

Sensitive data, privacy, and production stability are treated as constraints, not afterthoughts.

05
Reliable software beats AI gimmicks

A system that works every time is worth more than a flashy demo that works once. Reliability wins.

Project Intake

Have a workflow that should be software?

Describe what you do manually today. You'll get a straight answer about whether it's worth building — and a rough shape of the system if it is.

  • No sales pitch — a technical read on your problem.
  • A rough module / data / API shape, like the demo above.
  • Honest scope: what's worth automating, what isn't.
typical reply

"That's 3 entities, 5 endpoints, and one AI step worth adding. Phase 1 stands alone in ~2 weeks."

UI demo — validation runs client-side. Wire to POST /api/intake to go live.