I build practical AI systems
for teams that need faster operations.
My work sits between product engineering and delivery ops: reporting pipelines, proposal workflows, marketplace research, Microsoft 365 integrations, and full-stack apps that remove manual admin work.
Operational clarity
Map messy handoffs into one usable workflow before adding more software.
Systems that connect
APIs, reporting, automation, and human review points working as one operating layer.
Shippable scope
Start with the smallest version that can reduce admin load or speed up delivery.
System preview
Workflow mesh
Human checkpoints kept inside the system, not bolted on after shipping.
APIs, documents, and reporting outputs tied back to the same workflow state.
Interfaces shaped around what reduces admin friction fastest.
Applied AI where it helps
Useful classification, summarization, and review support instead of novelty features with no owner.
Workflow-first delivery
Clarify the system, narrow the first win, then expand only when the operating value is real.
What I Actually Help With
The strongest part of the portfolio is not an endless tech list. It is showing the kinds of systems you can trust me to design, build, and improve.
Client Operations Systems
Operational software that turns messy processes into repeatable workflows
- ▸Lead intake, proposal support, and inbox triage
- ▸Weekly hours, spend, and financial reporting exports
- ▸Workflow checkpoints for small teams and agency-style delivery
- ▸CLI-first tooling when speed matters more than UI
- ▸Desktop utilities when teams need repeatable local workflows
AI-Assisted Automation
Applied AI where it reduces manual work instead of adding novelty
- ▸OpenAI-powered message summarization and action extraction
- ▸Prompting patterns tuned for internal tooling and operator workflows
- ▸Research pipelines that turn raw marketplace data into useful reports
- ▸Document and communication analysis for faster review cycles
- ▸AI features scoped around reliability, reviewability, and human handoff
Integrations & APIs
Bridging third-party systems into one usable operating layer
- ▸Upwork GraphQL and REST integration
- ▸Microsoft Graph sync for mail and Teams data
- ▸OAuth and delegated auth flows for local tools
- ▸Encrypted token caching and privacy-first defaults
- ▸Structured exports in JSON, CSV, and Markdown
Product Delivery
Shipping working products across backend, frontend, and desktop surfaces
- ▸Next.js and React interfaces for public-facing and internal products
- ▸Python backends, scripts, and support services
- ▸PyQt6 apps for teams that still need desktop-native workflows
- ▸Documentation-heavy delivery so the system can be operated after handoff
- ▸Progressive implementation from rough idea to stable workflow
Working Style
Clear communication, realistic scope, and bias toward maintainable systems
- ▸Translate rough business requests into technical shape quickly
- ▸Prefer narrow, useful wins over giant rewrites
- ▸Keep reporting, acceptance criteria, and docs close to the code
- ▸Build with operators and non-technical users in mind
- ▸Avoid over-engineering until the workflow proves itself
Core Stack
The tools I reach for most often when building and refining systems
- ▸Python, FastAPI, and automation scripts
- ▸React, Next.js, and TypeScript
- ▸OpenAI integrations and structured prompt flows
- ▸Markdown-driven reporting and operational docs
- ▸GitHub-based iteration, review, and delivery
Typical Deliverables
Writing, Notes, and Perspective
The writing section works best when it reinforces how I think: practical engineering tradeoffs, workflow design, and the occasional personal context that explains why remote-first work matters to me.
Exploratory Feature Development: Validating Ideas Through Progressive Implementation
A practical write-up on validating workflow ideas by moving from CLI to GUI to broader application surfaces.
The Future of AI-Powered SaaS Applications
Where AI helps in real products, where it turns into noise, and how to choose features that survive contact with users.
Vector Databases: The Foundation of Modern AI
A grounded explanation of where retrieval systems fit into production software and how to think about them practically.
Living with Migraine: A Developer's Journey to Remote Work and Advocacy
A more personal piece on remote work, health constraints, and how lived experience shaped the way I work and build.
Selected Work
These are stronger portfolio stories than a generic list of stacked buzzwords. Each one shows a system built around an actual workflow, a practical constraint, and a clear use case.
Upwork Client Operations Platform
A working operations layer for job discovery, proposal support, inbox review, time reporting, and marketplace research inside one Python-first system.
What shipped
- ✓Job feed search with suitability scoring
- ✓Proposal and applicant export workflows
- ✓Weekly spend and hour reporting
- ✓Inbox summaries and follow-up support
- ✓CLI and GUI paths for different operator needs
Technology stack
Why it matters
Freelancer Research & Rate Benchmarking
A market intelligence workflow that searches freelancer profiles, groups results by experience tier, and exports usable pricing research in JSON, CSV, and Markdown.
What shipped
- ✓Preset and custom skill searches
- ✓Rate filters and experience banding
- ✓Top competitor summaries
- ✓Readable stakeholder-facing output
- ✓Retry logic around API limits
Technology stack
Why it matters
Microsoft Graph Local Connector
A local connector for Microsoft 365 data that syncs Teams chats and mail using delegated auth, encrypted token caching, and privacy-first defaults.
What shipped
- ✓Device code flow authentication
- ✓Encrypted token cache
- ✓Mail and Teams synchronization
- ✓Metadata-first exports
- ✓Tested client and auth modules
Technology stack
Why it matters
How I Usually Add Value
Most of the useful work lands in one of these buckets: shaping the workflow, building the first useful version, or tightening an existing system so it becomes easier to operate.
Audit
Review the current product, content, or workflow and identify what is actually worth changing first.
Prototype
Build a narrow, useful version of the system so the team can validate the direction quickly.
Integrate
Connect tools, APIs, and data sources so work can move through one cleaner process.
Stabilize
Reduce friction in an existing codebase or internal tool and make handoff easier.
Common Questions
This section does more work when it answers how I operate, what I build, and where I am useful, instead of repeating generic agency copy.
What kinds of projects are the best fit?▼
The best fit is usually an internal tool, workflow automation project, reporting system, or narrow product surface where there is real operational friction and a clear owner. I'm especially useful when a small team needs one person who can work across product thinking, backend automation, integrations, and frontend delivery.
How do you approach AI features?▼
Do you only build web apps?▼
How do you work with messy or changing requirements?▼
Can you work inside an existing team or codebase?▼
What does your process usually look like?▼
Do you write documentation and handoff notes?▼
Can you help with audits or cleanup before new features?▼
How do you handle privacy and sensitive data?▼
What should a first message include?▼
Still have questions?
Send a rough summary of the workflow or project and I can tell you quickly whether it is a good fit.
Start With The Real Workflow Problem
If you already know where the friction is, that is enough to get started. I can help shape the technical direction from there.
Send the actual brief
This form posts to a server-side route, then forwards the message to regan@rbtrends.com.au. Your email is set as the reply-to address so the response can come straight back to you.
What happens after you hit send
1. The message lands in one inbox
The API route forwards your note to regan@rbtrends.com.au, so the site is not relying on the visitor's local mail app opening correctly.
2. Your reply path is preserved
The sender email becomes the reply-to address, which means replies can go straight back to you without copying details out of a form submission dashboard.
3. The form is filtered and validated
Basic server-side validation rejects empty or malformed requests, and a hidden honeypot field helps drop obvious bot traffic.
4. The brief is the starting point
A rough workflow description is enough. The first conversation is usually about constraints, systems involved, and the smallest useful version to ship.
Useful starting points
- We have too much manual admin around one workflow.
- We need reporting people can actually use.
- We want AI support, but only where it is genuinely useful.
- This process is split across email, chat, spreadsheets, and separate tools.
Better portfolios usually say less and prove more.
That same rule applies to client work. Start with the real problem, make the first useful version small, and build trust through what actually ships.