Case studies, developer stories, and real-world software we've designed and deployed — from AI API integrations (OpenAI, Azure) to custom invoice generators, restaurant order pages, and AWS-hosted web apps.
Our client was a freelance iOS developer receiving payments from two sources every month: Stripe India sending INR payouts via Standard Chartered Bank, and Apple Inc sending USD royalties later converted by HDFC. Every month, the process was manual: download the bank PDFs, read the numbers, open a template, type the invoice, save the PDF, repeat for both clients, then calculate GST totals.
It worked — until it didn't. With growing volumes and compliance requirements, manual entry meant errors. A missed decimal place on a GST invoice could cause real problems at tax time. The ask was simple: automate everything.
"I receive bank advice PDFs every month from two clients. I want to upload them, have the system read the amounts, and generate compliant GST invoices automatically — with a shared invoice number sequence."
Using pypdf we built regex-based extractors that parse two very different PDF formats — Standard Chartered Bank payee advice (for Stripe INR) and HDFC bank remittance advices (for Apple USD). The extractor pulls invoice totals, dates, USD amounts, and exchange rates from unstructured text.
The system handles two separate workflows with a shared sequential invoice number (e.g., Apple gets #0025, Stripe gets #0026). ReportLab generates A4 PDFs with all required GST fields: GSTIN, seller address, invoice date, due date (date + 2 days), and itemised amounts.
Every invoice is automatically placed in the right folder: months/FY/Month/outputInvoice/ for Stripe, months/FY/Month/appleInvoice/ for Apple. At the end of each month, one click generates a monthly GST summary PDF showing all invoices and the combined INR total for tax filing.
A single-page HTML app running on Flask port 5050 gives the client a drag-and-drop interface to upload PDFs, view invoice history, regenerate invoices, and download everything. No command line, no spreadsheets.
The client now processes a full month of invoices in under two minutes. The system runs locally, requires no database, and persists data in lightweight JSON files. It handles Indian Financial Year logic (April–March), auto-increments sequences across clients, and cleans up stale records automatically.
Python 3 + Flask · pypdf for extraction · ReportLab for PDF generation · Vanilla HTML/CSS/JS frontend · JSON for persistence · Deployed on AWS EC2 / S3 — no database, no ORM, no framework overhead.
If you have a manual, repetitive workflow — whether it's invoicing, report generation, data extraction, or business automation — we can turn it into a clean web app. Tell us what you need.
Case Study #2
No POS system. No Zomato commission. No phone juggling. Just a single mobile-friendly menu page where customers tap items, fill in their details, and tap "Place Order" — the owner gets an email with the full order instantly.
Build One for My Business →A small family restaurant in a busy locality was taking all orders over the phone. During lunch and dinner rush, the owner's phone never stopped ringing — customers calling to ask "what's on the menu?", repeated orders being misheard, items going out of stock mid-service, and no record of what was ordered. Every evening ended with a stack of scribbled notes.
They looked at Zomato and Swiggy — but the 25–30% commission on every order made it unviable for a small kitchen. They just needed something simple: a digital menu customers could browse, and a way to receive orders without a phone call.
"I want a page where my customers can see the menu, tap what they want, enter their name and table number, and submit. I want to get an email with the order. That's it. No apps, no subscriptions, no commission."
One HTML file. Menu items organised by category (Starters, Mains, Breads, Drinks, Desserts). Each item shows a name, short description, and price. Customers tap + to add to cart, − to remove. The cart total updates live at the bottom of the page without any page reload — pure JavaScript, no framework.
Below the menu, a short form collects the customer's name, mobile number, table number or delivery address, and any special notes. On hitting "Place Order", the cart contents are compiled into a clean order summary and submitted via Formspree — the owner instantly receives a formatted email listing every item, quantity, and the total.
We generated a QR code linking to the page. The owner printed it and stuck it on each table. Customers scan it with their phone camera — no app install, no login. The page loads in under 2 seconds on a basic 4G connection.
The menu items are stored in a simple JavaScript array at the top of the file. The owner can edit item names, prices, and availability by changing a handful of lines — no database, no CMS, no admin panel needed. We showed him once; he's been updating it himself ever since.
Within the first week the owner stopped taking phone orders entirely. Customers preferred scanning the QR code — they could browse at their own pace, see prices clearly, and submit without waiting. The owner gets a clean email for every order and can see the full order history in his inbox.
The same setup works for grocery stores, bakeries, cloud kitchens, juice bars, and any small business that wants to take orders without paying platform commissions.
HTML + CSS + Vanilla JavaScript — single file, no framework · Formspree for order emails (free tier: 50 orders/month; paid: unlimited) · Static hosting on GitHub Pages or any web host — costs as little as ₹0/month · QR code generated with any free QR tool.
Restaurants · Dhabas · Cloud Kitchens · Grocery Stores · Bakeries · Juice Bars · Sweet Shops · Tiffin Services · Catering Businesses · Pharmacies
We can build a custom order page for your restaurant or shop in under 2 weeks. You choose the menu categories, items, and price. We build it, deploy it, and hand it over — you own it completely, no ongoing fees. Fill in your requirements here.
AI API Integrations
We build custom software around the most powerful AI APIs available — OpenAI (GPT, Whisper, DALL·E, TTS), Microsoft Azure (Speech, Vision, Cognitive Services), Google Cloud AI, and more. If you have an idea that needs AI — text, voice, image, or document intelligence — we can build it, wrap it in a clean interface, and deploy it on AWS.
Custom TTS apps using OpenAI TTS, Azure Speech, or ElevenLabs. Realistic voices, multiple languages, export to MP3/WAV. We've shipped iOS apps with 24K+ users doing exactly this.
OpenAI TTS · Azure Speech · ElevenLabsTranscription tools, meeting recorders, voice-to-form apps. Uses OpenAI Whisper or Azure Speech SDK. Handles accents, background noise, and multiple speakers. Works on web and iOS.
OpenAI Whisper · Azure STT · Apple STTExtract text from images, analyse product photos, detect objects, generate image descriptions, read receipts and documents. Using OpenAI GPT-4o Vision or Azure Computer Vision.
GPT-4o Vision · Azure Computer VisionCustom chatbots, document Q&A, content generation tools, email drafters, or any app that needs a language model under the hood. We wire GPT-4o or Claude to your data and workflow.
OpenAI GPT-4o · Claude API · RAGEvery project we build can be deployed on AWS — EC2 for apps, S3 + CloudFront for static sites, RDS for databases, Lambda for serverless. We handle infrastructure so you get a production-ready URL from day one.
EC2 · S3 · CloudFront · Lambda · RDSTell us the problem. We'll figure out the right API, build the integration, and deploy it for you.
Start a Project →You describe what you want. We choose the right API (OpenAI, Azure, or other), build a clean interface around it, and deploy the whole thing on AWS. You get a working URL, source code, and documentation. No vendor lock-in, no ongoing agency fees — you own everything.
All Articles
How a simple one-page website replaced phone orders for a local restaurant. Customers scan a QR code, browse the menu, and submit their order — owner gets an email instantly.
Small BusinessA neighbourhood grocery store was managing orders through a WhatsApp group. Chaos. We built a single webpage with categories, quantities, and a submit button — orders go straight to email.
Small BusinessA home catering business had no web presence. We built a one-page site with a menu showcase and enquiry form — name, event date, guest count, menu preferences. Owner gets an email for every enquiry.
AI APIA content team needed a web tool to turn scripts into MP3s using OpenAI's TTS API. We built a clean single-page app — paste text, pick voice, download audio. Deployed on AWS S3 + CloudFront.
AI APIUpload an audio file, get a full transcript and summary back in seconds. Built with OpenAI Whisper + GPT-4o. Runs on AWS Lambda — pay only when someone uses it.
AI APIUpload a photo of any receipt, invoice, or document — the app extracts vendor, date, line items, and totals into a structured table. Built with GPT-4o Vision, deployed on AWS.
AI AppsBehind-the-build look at Ondevtra's AI Voice Generator — multiple voices, languages, speed controls, and one-tap MP3/WAV export.
AI AppsHow we built an AI stylist that turns product photos into on-model looks — helping users visualise outfits before buying.
AI AppsBuilding a study companion that converts voice scripts and subtitles into quiz questions and flashcards for content creators.
Custom SoftwareHow AI-powered custom software — from invoice generators to reporting dashboards — is cutting hours of manual work each month.
AI AppsThe engineering decisions behind building a low-latency speech-to-text app that works across accents and noisy environments.
AI AppsDesigning a low-latency translation pipeline for live conversation — challenges, architecture decisions, and lessons learned.
Web DevThe frameworks, deployment strategies, and tooling choices that define production web development today — and what to avoid.
Web DevFrom CSS layers to island architecture — the frontend patterns worth learning and the hype cycles worth skipping.
AICutting through the noise — which generative AI capabilities have real ROI and which are still science experiments.
AIThe prompting techniques that consistently get better outputs — including chain-of-thought, system prompts, and constraint setting.
VLSIA structured guide to physical design flow — from floorplanning and placement through CTS, routing, and sign-off.
iOS DevThe Swift and SwiftUI patterns that survive contact with real App Store submissions — async/await, Combine, and state management.