Backend Developer & API Integrations
Cloudflare Workers, data, realtime systems and external APIs
I build the backend logic that connects products to their data and providers. My project work includes Cloudflare Worker APIs, Durable Objects, WebSockets, SQL data, scheduled queues and integrations for commerce, payments and realtime audio. My wider programming experience includes C#/.NET, Python and PHP.
Five years of hands-on programming before adopting AI. That is my pre-AI foundation, not my total experience. Today I use AI as a development tool while retaining responsibility for architecture, understanding the code, validation and delivery.
How I approach this work
- Model account, payment and inventory records around explicit rules and change history.
- Handle external events, duplicate payment notifications and scheduled work without tying every operation to an interactive request.
- Investigate provider errors, processing costs and latency across the frontend, backend and integration boundary.
Skills & technical capabilities
Systems
Reliable logic, data and integrations behind the product.
- Application logic
- C# / .NET · Python · PHP · Backend architecture
- Data & APIs
- SQL · Data modeling · Databases · REST · APIs
- Infrastructure
- Cloudflare Workers · Integrations · Authentication · Caching
Project experience
Realtime translation
Viva Talk
Live speech translation for browser conversations, with a separate Zoom integration.
Technologies: Next.js, TypeScript, WebSockets, Durable Objects, Zoom RTMS, D1, AI Integration
Public product link and project screenshots.
Conversation comes first
Viva Talk supports 25 languages and up to 5 concurrent language pairs in browser conversations. Participants join a room, choose their language and hear one another in that language. My work spans the meeting interface, audio transport, translation backend, account access, billing and an admin console. Zoom support is a separate beta integration with its own audio receiver.
Audio, state and billing
- Room state lives in a Durable Object; WebSockets carry participant events and audio.
- Listeners who share a target language share a translation session. Those who speak the same language receive the original audio without AI processing.
- Account credits and session charges are recorded in a ledger, with duplicate payment events handled explicitly.
- Initial translation tests measured approximately 0.5–1.5 seconds, with occasional longer delays. Latency, provider errors and estimated processing costs are recorded separately.
Keeping meetings usable
The challenge was to bring realtime speech translation, conversation summaries, translated chat messages, screen sharing and video into one intuitive meeting experience. These features needed to work together smoothly, so participants could focus on understanding one another.
Robot operations platform
RoboMatrix
Customer and admin workspaces for trading robot accounts, balances, payments and support.
Technologies: Next.js, React, TypeScript, Workers, D1, R2, APIs, Crypto Services
Private product; portfolio demonstration uses sample data.
One product, two sides
RoboMatrix combines 2 customer/admin workspaces for robot accounts, packages, balances and support. As the sole software developer, my work spans the React interfaces, Worker APIs, data model and external account integrations.
Robots, partners and operations
- 10+ robot types and 5 packages connect to account settings, statistics and operational controls.
- Server integrations bring robot information and statistics into the customer and admin workspaces.
- A partner program with 6 ranks covers referral structures, partner statistics and compensation.
- USDT and USDC payments are supported; bank, prepaid and payment activity have separate records.
- The backend handles package activation, refunds and partner compensation through dedicated services and ledger entries.
- SpendX account integration, authentication, notifications and support sit behind the same application boundary.
What is shown here
This work required translating account and payment rules into interfaces that customers and operators could use. The portfolio preview is a demonstration with sample data; it does not expose the private application or live accounts. The focus of this example is the account platform and its operational workflows.
Operations platform
Mixar Control
A private commerce workspace connecting stock, orders, shipping, receipts and customer messages.
Technologies: Next.js, TypeScript, Workers, D1, Commerce APIs, PHP, AI integrations
Private product; portfolio demonstration uses sample data.
The work between systems
Mixar connects 9 commerce services for a store with 500+ products, including SalesDrive orders, Nova Post shipments, Checkbox receipts and customer messaging. My work covers the operations interface, Worker backend and integration code. The database also holds nearly 55,000 delivery service points.
Integration details
- Stock changes retain their order reference and change history; product edits check the expected modification time before overwriting content.
- 3 scheduled workflows handle messages, order status changes and inventory synchronisation outside the interactive request.
- An OpenCart add-on for a separate storefront synchronises products.
- Operators can work with product content, shipment confirmation and receipt workflows from the same workspace.
- An internal AI assistant can investigate stored data and queue code tasks. Database actions record a recovery bookmark, and code tasks are restricted to the project’s repositories.
Built around the operator
Automation of orders, inventory, shipping and customer notifications reduced manual work by approximately 70% (operational estimate). The portfolio preview uses sample data to show the workspace; the private integrations and customer records are not connected to it.
Engineering approach
I work across product boundaries: interface behavior, application rules, data models and integrations. I structure systems so that the next developer can understand and extend them, and investigate failures across the complete flow.
- Design across the complete flow, from user action to backend state and external provider response.
- Keep interfaces, business rules, content and integrations separated so that changes remain understandable.
- Validate behavior in the browser and on the relevant platform; distinguish previews and sample data from live product behavior.
- Use AI to assist development while reviewing the code and taking responsibility for the result.
Contact & next step
To discuss a role, use the portfolio contact form and choose Hiring. Include the role, product context and the experience you would like to explore.