Technical partner to a US PDR company since 2021: server, telephony, super app and document parsers

Since 2021 I have been the technical partner to a PDR company in the US. PDR stands for paintless dent repair. I own the server and network, the custom API, telephony, the document parsers, the TV board and the leaderboard; the React Native super app we build together with the company's developers. The main result: thanks to automation, lead flow grows several times over from year to year, while the sales team has grown only a little.

I don't disclose the company's name, city or figures.

The project in brief

• Industry and period: PDR, US; since 2021, ongoing.
• My role: full-stack developer and DevOps, architect, product owner of my projects, mobile app developer.
• The company employs more than 30 people today.
• Stack: Debian on Google Cloud, MariaDB and SQL, Node.js, Next.js and EJS, React Native, AppSheet, in-house PBX and IVR, Twilio, GoHighLevel, Python with OpenCV, Tesseract, Camelot and Pandas, Puppeteer, AI services from Anthropic, OpenAI and Gemini.

What the company got

The company used to have 20 employees; today it has more than 30, and over that time lead flow has grown several times over, year after year. The sales team has grown only a little and still handles every inquiry: each salesperson gets more leads every year, but the load per person doesn't multiply. Processing got faster and better. Every real lead goes all the way through the funnel, and conversion to sale is up 30%.

The system records every call, and AI works out who called, a lead, an existing customer or an insurer, and how the employee handled it. We always know who called, when and why.

Parsers read insurance estimates, work certificates and VIN photos: about 99.9% of documents go through without manual correction, which frees up about 60 hours of work a week.

Since 2022, when the company got its own server, there has been no serious outage that needed a recovery.

How I joined

In 2021 the company's owner found me on Upwork himself, by my reviews and rating. The company put together documents for repairs, for insurers and for customers. Data from several sources had to come together in one PDF, go out by email and straight to the office printer; then the document was signed and filed.

I wrote the logic myself; the printer and the rest of the physical setup we did with the office staff, since the company is in the US and I'm not. The owner liked the work and offered an ongoing engagement: he said he wanted strong specialists to stay with him for a long time. That's how it went: we still work together, and I like it here.

Before and after

In 2021 the data lived in Google Sheets, the second developer built AppSheet apps on top of them, and the sheets slowed everything down. The website sat on shared hosting, nobody handled network security, and staff typed data from documents into a spreadsheet by hand.

Today the data sits in a SQL database on the company's own server, next to the API, the website and telephony on an in-house PBX with AI call analysis. The center of the system is the super app for salespeople, front office, back office and customers, and GoHighLevel went from the main CRM to a supporting one. Team metrics show on the leaderboard and the TV board, parsers read the documents, and a person does the final check.

How a lead moves through the system

Website, forms and IVR calls → database and super app → salesperson in the app → site visit, working on location offline first → documents generated automatically → CRM, TV board and leaderboard

• An IVR call: the system creates the lead in the CRM on its own and attaches the recording and the AI breakdown of the conversation.
• The salesperson sees the request in the super app and calls and texts the customer from there. On site they drop a pin on the map, take photos and work the lead without a connection; the app syncs the data once it's back online.
• The system assembles estimates, work certificates and invoices into PDFs from its own data. Incoming insurance documents of 60 to 70 pages go to the parsers.
• Metrics by employee and by job flow to the leaderboard and the TV screens.

My role and the team

We agree on projects as a group of three: the owner, the company's second full-time developer and me. After that each project has its own owner. The second developer runs AppSheet, where he has built a lot of apps. I own the backend, server, network and security, databases, VPN, telephony, the internal API, the website, the TV board, the leaderboard and the business cards. Product owner, in my case, means I'm responsible for a project end to end, from idea to support.

Many of the ideas were mine: move to a SQL database, set up our own server, build a PBX with AI call analysis. The owner accepted them, and we changed how the company works accordingly; now he does development with us himself.

Projects in order, 2021 to 2026

2021. Automatic PDFs

The task is described above. I still run PDF generation, now on Puppeteer: work certificates, estimates, reports, invoices and agreements.

2022. Server, VIN and the CRM integration

We took on decoding VINs and documents, and to tie it all together I set up a new Debian server on Google Cloud. From that server I started rebuilding how the company handles data. I built the VIN and document decoding. We did not do license plate decoding, which would pull owner and vehicle value data: the external service cost money, although it was doable. The same year I wrote the integration between the salespeople's AppSheet app and GoHighLevel. The company also built a hiring app that year; it used the app for about a year, then moved to a different approach. On the server: Debian on Google Cloud.

2023. SQL database, website on our own server and the first parsers

I proposed moving from Google Sheets to a SQL database for speed and a single data format. The owner backed it, and after the move to MariaDB we connected new services one after another. I kept extending the GoHighLevel integration all year. I moved the company website from shared hosting to our server and connected its forms to the CRM and the app. That year I also started the document parsers: they have been in use since mid-2023, when about half of the documents went through without manual correction. In detail: Document and photo parsers. On the CRM: GoHighLevel.

2024. Native app, cards and QR, the start of IVR

The move to SQL gave us momentum: we started building the first native app for internal use. My projects that year, from idea to release: virtual business cards for employees, from which you can download a vCard, call or message, and a QR code generator with short links to them. I also started the IVR system that year, and the leaderboard at the end of it.

2025. TV board on 4 screens and an IVR interface

I put the TV board on 4 TVs in the company, each with its own board: for the owner, for production by job, for the front desk and for salespeople. I built a graphical interface for the IVR. I developed the leaderboard all year: leads and sales by manager, percentages, competitions and team splits. On the leaderboard and TV: Calls, CRM and service team workflows.

2026. PBX with AI, parser release, super app

I took the IVR to a full in-house PBX with custom logic; it pushes the relevant information to the desk phones. I connected AI validators to the parsers and shipped the release version: about 99.9% of documents without manual correction. Over these years I took the internal API to its third version: A custom API for connected applications.

The React Native super app we build together with the company's developers; it is replacing AppSheet, and the office, the salespeople and the owner already use it. The app is on TestFlight, the App Store release is in preparation, and for Android we build an APK. It's built offline first, and salespeople use it right on site: map, photos, leads. Next to it is the customer portal: an online timeline for the job, signing a document with a finger with a PDF generated from it, photos, push and SMS notifications, in English and Spanish. I'll write separate pages on telephony, the super app, the customer portal, cards and QR.

Security and resilience

Security questions in the company come to me. The system has several layers of protection and backups. I monitor the infrastructure, and when a major breach hits the news I check whether it concerns us. I maintain the server, VPN, network equipment and telephony myself.

What it gave me

Over these years I've grown as a specialist: from individual automations to a system in which the server, telephony, the app and the documents work as one. I've gained experience in the US market, with American clients, and in real development for the automotive field.

A similar task?

If your company's data is scattered across spreadsheets and services, calls get lost and documents are retyped by hand, tell me how the work runs today and where it gets stuck. I'll tell you where to start. No need to send confidential documents; a description is enough. Discuss a similar task.

By area: CRM integrations, internal systems, API integrations, DevOps and infrastructure, IP telephony, AI workflows.