# Rev Insights
LLM chat for research and legal customers that allows users to get useful insights from their files. It lives inside the rev.com ecosystem so they can pull files from other modules like transcripts, meetings, depositions, and more.


 <img src="/images/projects/rev-insights/01.png" alt="drawing" width="700"/>
 Empty chat


## Why did we build this?

Customers already had a lot of content inside Rev — transcripts, meeting recordings, depositions, and uploaded documents — but no simple way to ask questions across that material. Insights gives them a chat interface that can pull those files in context, so they can search, summarize, and dig into evidence without leaving the product or stitching tools together by hand.

We also needed a competitive answer to the wave of standalone LLM chat products. Legal and research buyers were starting to expect AI-assisted workflows, and without a first-party experience inside Rev we risked losing that engagement to external tools that did not understand our content model or customer workflows.


 <img src="/images/projects/rev-insights/02.png" alt="drawing" width="700"/>
 Document Editor


## My participation

Rev already has a strong engineering structure, so this was a very typical engineering team. We had a UX designer, a PM, a tech lead, and 3 senior engineers. I had the role of senior engineer and participated in all aspects of the product.

## The tech stack

It runs mainly on Node + React + TypeScript.

Front-end

- React 18 + TypeScript
- React Router
- Apollo Client + GraphQL
- graphql-codegen
- Zustand
- Tailwind CSS v4
- react-markdown + remark/rehype
- SSE streaming
- Uppy + AWS S3

Backend

- .NET
- YARP
- GraphQL
- SSE
- Amazon SNS
- Amazon SQS

Testing & tooling

- Jest + Testing Library
- Playwright

LLM chat for research and legal customers

Rev Insights

# Rev Meeting Hub
Web tool for scheduling bots and recording meetings with AI.


 <img src="/images/projects/rev-meeting-hub/01.png" alt="drawing" width="700"/>
 Meeting Hub home page


## Why did we build this?

Rev saw a growing business opportunity in AI meeting tools. Customers already trusted Rev for high-quality transcription, but they still had to manually record calls, upload files, and wait for results. Meeting Hub closes that gap by sending an AI notetaker into Zoom, Google Meet, and similar platforms, capturing the conversation live, and feeding it straight into Rev’s transcription pipeline.

By building on Rev’s existing speech AI instead of stitching together third-party recorders and transcription vendors, we could offer a tighter product: automatic scheduling from connected calendars, reliable capture, searchable transcripts, and a meeting workspace that sits naturally inside the Rev dashboard. That made Meeting Hub both a new growth surface and a way to deepen engagement with customers who already used Rev for transcripts.


 <img src="/images/projects/rev-meeting-hub/02.png" alt="drawing" width="700"/>
 Transcript editor


## My participation

Rev already has a strong engineering structure, so this was a very typical engineering team. We had a UX designer, a PM, a tech lead, and 3 senior engineers. I had the role of senior engineer and participated in all aspects of the product, from backend to front-end, including developing and maintaining a microservice for Meeting Hub data.

## The tech stack

It runs on Node + React + TypeScript.

Front-end

- React 18 + TypeScript
- React Router
- Apollo Client + GraphQL
- graphql-codegen
- React context
- MUI + Emotion
- GrowthBook

Backend

- GraphQL
- .NET
- Google and Outlook calendar OAuth

Testing & tooling

- Jest + Testing Library
- Playwright
- ESLint + Prettier

Web tool for scheduling bots and recording meetings with AI

Meeting Hub

# AI Construction ERP
AI assistant for [Construction ERP](/project/hermosillo-construction-erp). It helps construction teams get information without digging through modules, learn how to use the system, cross data with tools like SAP and Autodesk, and generate custom reports.


 <img src="/images/projects/ai-construction-erp/01.png" alt="drawing" width="400"/>
 Simple question



## Why did we build this?

Construction ERP holds a lot of operational data: projects, procurement, estimates, billing, checklists, Autodesk files, and more. People still needed a faster way to get answers and to combine that information with other tools they already use, like SAP, Autodesk Forma and OPUS (Budgeting software).

Opening several modules, exporting spreadsheets, and reconciling numbers by hand was slow. Documentation existed, but it was hard to find at the moment someone was stuck. We needed an assistant that could:

- Answer questions about projects, purchase orders, estimates, billing, users, and related records.
- Help people use the system by pointing them to documentation and the right workflow.
- Cross data across Construction ERP and external platforms such as SAP and Autodesk.
- Generate custom reports when a standard screen was not enough.

The first version of this product was a chat bot. It started on Google Chat with slash commands and structured cards for things like project lookup (`/obra`) and purchase orders (`/oc`). That worked for a small set of questions, but it could not keep up with the way people actually ask for information.

We rebuilt it as an AI assistant. The model talks to Construction ERP through MCP tools, respects role and project permissions, and asks for confirmation before write actions. The same backend now serves Google Chat and the assistant inside Construction ERP web and mobile, with conversation memory, streaming answers, and usage reports so we can see what people actually ask for.


 <img src="/images/projects/ai-construction-erp/02.png" alt="drawing" width="400"/>
 Documentation


## My participation

I am the original author and main contributor of this project, and I designed the architecture.

I defined how the assistant connects to Construction ERP: the MCP tool layer, permission checks by role, write confirmations, conversation persistence, and the shared API used by Google Chat and the Construction ERP apps.

As the product grew I kept owning the core design while other developers contributed tools, prompts, and product features on top of that foundation.

## The tech stack

For this project we used Node on the backend and an LLM connected to Construction ERP through MCP.

- Assistant
 - LLM via the Open router
 - MCP client and tools against Construction ERP, SAP, and Autodesk
 - Role and project permissions
 - Write-action confirmations
 - Conversation memory (Sequelize / MySQL)
 - SSE streaming for Construction ERP web and mobile
- Channels
 - Google Chat
 - Construction ERP web and mobile
 - Local CLI for development
- Backend
 - Node.js
 - Express
 - Botkit
 - Jest
 - Internal npm packages
- Other
 - Usage analysis and HTML adoption reports (Puppeteer)
 - Docker / PM2 for deploy
 - GitHub Actions for CI/CD

AI assistant for Construction ERP that helps people find information, use the system, cross data with internal + external systems, and generate custom responses.

AI Assistance

# Construction ERP
ERP for the construction industry. It has modules for estimation, procurement, site management, accounting, and executive reports.


 <img src="/images/projects/hermosillo-construction-erp/01.png" alt="drawing" width="700"/>
 Home page of Construction ERP


## Why did we build this?

As a construction company with more than 50 years of experience we needed to take the next step into digital transformation, so we bought an ERP software from a germany company that was supposed to solve our problems. We tried to implement it for 2 years. At the end the implementation failed due to the following reasons.
 - Cultural differences in the software, process were not similar to Mexico.
 - Changes in the platform took a lot of time and were buggy.
 - We were the only client in LATAM so it was not worth it for them to maintain these changes.

During the implementations I started developing small pieces of software within the ERP to help solving the issues we were having using FastReport(C#). These programs helped a lot with the implementation but the dev process was really slow since I had to come up with a lot of workarounds to access the data on the ERP. I had to write very complex queries in SQL Server to perform heavy math calculations, I was saving data in JSON files and more.

After many tries I convinced my boss and the board that we could build a much better product if we built it on our own. Due to the nature of the project they let me build a prototype for one of the modules, it was a small module to handle the procurement schedule of a site project. The users were very happy with this new approach, they could access this module using their browser instead of running a heavy software via remote desktop, it was faster and more friendly.


 <img src="/images/projects/hermosillo-construction-erp/02.png" alt="drawing" width="700"/>
 Menu of all the modules


## My participation

This is the biggest project I have been part of and where I have taken more responsibilities. Since it was developed inside a construction company and not a software company I had to work in several roles. I was product owner, project manager and tech lead.

At the beginning of the project it was only me. After a few months we hired a more experienced developer that would help us to build the platform quicker. When we had the core modules then it was time to work on all the details so we hired 3 junior devs to help us to achieve that. At the end we were 5 developers in the team.


 <img src="/images/projects/hermosillo-construction-erp/03.png" alt="drawing" width="700"/>
 Procurement control module


## The tech stack

For this project we used a wide variety of framework and tools. We mainly used Node and React.
 - Front-end
 - Firebase
 - Redux
 - React
 - SemanticUI
 - React router
 - Backend
 - Mysql
 - ExpresJS
 - Jest
 - Heavy math computation using Decimal.JS
 - Sendgrid
 - SAP, Autodesk integrations
 - Other
 - We used microservices when we start facing DB issues and performance issues
 - We built a service in python to work generate pdf files.
 - We used docker for analytics
 - We built custom npm packages for components like Table, utils, logs, etc

ERP for construction process. it has modules for estimation, procurement, site management, accounting, and executive reports

Construction ERP

# Residencial Digital
Software to manage private residential neighborhoods — access control, fee payments, and more.


 <img src="/images/projects/residencial-digital/01.png" alt="drawing" width="700"/>
 Home page


# The project

In my city, private neighborhoods are very popular. I was first approached by one of them; they needed a software solution to manage all of their processes. I built it from scratch in three phases.

- Phase 1 was the development of a mobile app and a web app so the committee could manage information about houses, share messages, schedule events, and manage security personnel. Individual residents could manage their visitors and generate access codes.
- Phase 2 included a payment system. Residents can pay with Stripe or bank transfers, and the committee approves or rejects each payment. This phase also included reports for all payment information.
- Phase 3 connected the platform to a ZKAccess C3-400 access control system. We built the integration to activate or deactivate user access; residents who had not paid were disabled through a scheduled job. I also built reports for access using RFID tags.

I then offered this solution as a SaaS product to two other private neighborhoods.


 <img src="/images/projects/residencial-digital/02.png" alt="drawing" width="700"/>
 Payment page and Visit page


# My participation

I personally built the entire solution myself: the app, the API, and a Windows service that connects to the C3-400 controller for the access doors. I deployed it on a DigitalOcean VPS and used Spaces (S3-compatible) for media storage.

# The tech stack

- Frontend
 - React.js
 - Redux
- Backend
 - Node.js
 - Express.js
 - MySQL
- Windows Service (C3-400 connector)
 - .NET
 - NSSM
 - AccessDB
- Other
 - DigitalOcean VPS
 - Ubuntu server
 - Nginx

Software to manage private residential neighborhoods — access control, fee payments, and more.

Residencial Digital

# Meson
Search engine result page (SERP) that uses a custom framework built with Node and React


 <img src="/images/projects/amg-meson/01.png" alt="drawing" width="700"/>
 SERP


# The project
Ask Media Group main product is a SERP. The SERP is deployed and hosted by many domains, a domain may have some special features but 95% of the product is the same. They had developed a custom framework called Archie, it was built on top of Node.js. A quick summary of Archie, it had some core repositories for managing components and rendering using dust.js templates, for the front-end it had a lot of repositories that contained around 6-7 domains, so the more domains they have, the more repositories they need, built time was a problem and also the cost was increasing a lot. It got to a point where they need it to come up with a better framework, they started a new project called Meson.

Meson was the new framework to solve all the cost and performance problems. They also developed a new tool for business in Retool so they can perform small changes in the SERPs, like font-size, font-color, logo position and more, Meson had to take into account this configuration to render the final SERP. The whole project took around 1.5 years to be completed.


 <img src="/images/projects/amg-meson/02.png" alt="drawing" width="300"/>
 Mobile version of SERP


# My participation

The first phase of the project was to develop the core of Meson. It was composed of a custom Builder (similar to Next.js) and a Core engine. Both were built by 2 architects that had been working in AMG for more than 10 years.

The second phase of the project included the migration of all features and components from Archie to Meson. This is where I joined the project. I started as a regular senior developer then at the end I was named co-tech lead. My main task were
- I was in charge of migrating components from Archie to Meson.
- I was in charge of deploying new releases of Meson
- A / B testing for Meson and Archie in fastly
- Help and guide developers from different teams to understand how meson works.


 <img src="/images/projects/amg-meson/03.png" alt="drawing" width="700"/>
 Content page in non english language


# The tech stack

In this project the main framework was Node.js along with React. We used SSR using a custom framework similar to Next.js. We had a mono-repo where all the components were developed and tested. The full list is
- Frontend
 - React.js
 - Redux
 - Style Less
 - Dust.js
- Backend
 - Node.js
 - Redis
 - Express.js
- Other
 - Google, Yahoo and Bing APIs
 - Internal API for content
 - SSR
 - Fastly vcl

Search engine result page (SERP) that uses a custom framework built with Node and React

Meson

# SmartDepo
Web tool for generating deposition documents with AI, built with Python and React.


 <img src="/images/projects/rev-smartdepo/01.png" alt="drawing" width="700"/>
 SmartDepo home page


## Why did we build this?

Rev saw a clear opportunity in the legal market. Law firms already relied on Rev for deposition transcripts, but turning those transcripts into polished deposition documents was still slow, manual work. SmartDepo filled that gap with an AI-assisted workflow that can generate deposition documents from transcript content, cutting hours of formatting and drafting for legal teams.

Rather than building that capability from scratch, Rev acquired SmartDepo and brought it into the legal product suite. That let us move faster on a specialized workflow while tying it into Rev’s broader transcription and legal tooling, so customers could stay inside one ecosystem from recording through finished deposition deliverables.

## My participation

This was an unusual team setup. Two SmartDepo founders stayed on to help with the transition — one as PM and the other as a senior developer. My role was to add new features to the product. The main one I shipped was multi-witness detection, which lets the system support and produce depositions with more than one witness.

## The tech stack

Front-end

- React + TypeScript
- React Router
- React context providers — no Zustand/Redux
- MUI (@mui/material, Data Grid, Date Pickers) + Emotion
- GrowthBook (feature flags)

Backend

- Python + Flask
- AI services: Gemini + ChatGPT

Web tool for generating deposition documents with AI, built with Python and React

Julio Ojeda

Featured projects

Rev Insights

2026, February

Read more →

Meeting Hub

2024, January

Read more →

AI Assistance

2026, January

Read more →

Construction ERP

2020, January

Read more →

Residencial Digital

2021, January

Read more →

Meson

2023, January

Read more →

SmartDepo

2025, June

Read more →

Punto de Venta

2022, January

Read more →

Expedientes digitales

2018, January

Read more →

Geologistic ERP

2022, January

Read more →