Getting the Job by Setting a Vision

To interview at Rivian I needed to put together a vision deck that walked through the steps of what I would do if hired. This wasn't a technical interview, it was a panel presentation where the focus was on strategy and insight. The group I was joining was a collection of architecture, design and construction people. I was explained they wanted to hire someone who could set a vision and execute, and I was given the following prompt to design a deck and presentation:

At the time I was working at Autodesk in their Consulting Services (non-commission). My job was to help enterprise customers (large architecture and construction companies) build solutions on a new offering called Autodesk Platform Services (APS, formally known as Autodesk Forge). In a nutshell APS is two things: it's what they branded as their CRUD APIs for Autodesk SaaS platforms like Autodesk Construction Cloud. And second, what is less obvious, it's a set of tools for building new 3D applications by offloading the complexity of 3D to a set of web services that help render 3D files, translate file formats, create PDFs from views, and much more. I really enjoyed this job because it put me in front of people who were in charge of trying to innovate in their respective spaces and I got the opportunity to hear about what problems they were trying to solve.

Large architecture firms viewed this technology through the lens of their core business, which is a chain of workflows that ultimately leads to creating PDF files. A cynical me would say MOST businesses could fall into this category, but let's not dwindle there. Three main themes I ran across were workflow optimization, quality control and insight generation. Leadership at these firms wanted a quality product, but since this wasn't easy to measure, they found themselves focusing on the data they did have which was how profitable their projects were and used this KPI to measure technology initiatives.

Rivian was expanding their facilities footprint in 2022 and even though they are an owner, during this phase they emulate a design and build architecture firm. They employ dozens of project managers, architects, construction managers to manage 10+ projects each. The bulk of the work is done by external AEC vendors that get plans designed, approved by the city and built. During this phase they focus a lot on standards and prototypes trying to develop the brand and how it translates to the space look and feel. As they grow, their facilities footprint will grow and the goal is to bring down the $ / SQFT it costs to go from location selection to operational facility. Rivian wanted to hire someone who could look at their workflows and figure out if there was an opportunity to create tools and automations that could help save money.

If you look at any digital workforce,  I put together this diagram to think about the chain of events that need to happen and in what order:

The Connect as Needed was the job I was interviewing for. This panel was less interested in the how and more looking for insights on the what.  I chose 3 themes to present on, given my current perspective working for Autodesk Consulting:

🫙Collect and Store Structured Data (Datalake)

Often when consulting, someone I was helping would say something on the lines of “wouldn’t it be cool, if when a construction document was published we could send an email to all of the stakeholders automatically. without having to attach them manually.” If you were to break this problem down and get to what needed to be built to create an automation like this, you would need data from multiple sources. The Autodesk Construction Platform would have the data when the document set was published, but most organizations had their team structures in a different system, some came from Workday, some came from other internal systems that tracked the project lifecycle.

When’s the best time to plant a tree? Yesterday, and the next best time is today. When is the best time to start data warehousing when you have committed a long term on the platforms you chose? …

Data warehousing has been a buzz the last ten years but as a consultant on the frontlines of “wouldn’t it be cool” conversations I can vouch for the value it brings. By investing in storage, data engineering and identifying a team or individual who maintains these systems it opens the doors to a world where a company's automation goals can be met. And in addition, we haven’t talked about dashboarding yet, but it enables insight generation, but that’s a different topic I will touch upon a little later.

Rivian had committed to several platforms, one of them being Autodesk Construction Cloud. My hunch was if not today, tomorrow they will definitely need this data in a format accessible to automations and analysts. 

🔐Simplify Security

I love asking provocative questions, and my favorite question to ask is “How sensitive is this information, really?” In AEC, most building permits become publicly available. Unlike, for example product design, where the drawings and assets are only shared with the parties manufacturing the parts.

AEC is prime for a simplified security model which could lead to a lot of time saved for all parties involved. An example of a simplified security model is a construction document set that is accessible through a public URL that doesn’t require login (like sharing a Google Doc where the share setting is set to “Public”). Or for example access is granted based on a simple rule: if your company is part of a project you can be added automatically without approval as a viewer.

I would spend countless hours talking with an organization's cyber security team and their job, it feels like, is just to say NO. I sympathize because at the end of the day, when something goes wrong, the fingers point. If you start collecting a list of risks I think it’s important to add “company (department) is shut down due to inefficiency” or "lost moral due to slow tools". Someone needs to take a risk otherwise there is a general entropy to inefficiency.

The hallmark of any start-up is speed and that implies cutting corners. My vision of a thriving digital workforce is where information flows free. Where links in emails quickly provide insights and not frustration. Where you don’t wake up to another nightmare where your 2-factor authentication code wasn’t working. I attached my interview presentation below, you can see my suggestions on how to simplify security.

🫶Give Design Files a Second Life

Years of modeling, meetings and at the end of the project you have a set of PDFs, while the rest of the rich 3D design files are left on a drive somewhere never to be seen again. Autodesk has been fighting this up-hill battle for a while now but in the last five years a lot of progress has been made. But the innovation isn’t what we all expected.

Today it is a lot easier to share and explore our 3D and 2D assets outside of the authoring tools. Autodesk Platform Services Viewer is very powerful and enables bringing BIM (Building Information Models) in front of people who are typically viewing PDFs. There are two main challenges but if overcome, what a great opportunity to amplify our workforce.

The first challenge is 3D literacy. There are people in our workforce that feel comfortable navigating 3D space but a lot of jobs in AEC don’t require any 3D model interaction. For example if you are a cost estimator for a lighting company you typically get a set of 2D PDF files and you use an Excel Spreadsheet and your eyes to count and log how many lights are on a floor plan. If you offer someone a 3D model, oftentimes they aren’t sure how to use it to do their job. I had this theory that as the work force became younger more and more people would feel comfortable but I sat down with an avid gamer in their 20s and watched them struggle navigating a building 3D model.

The second challenge is legal risk. The construction documents are a legal document while the 3D model isn’t. Being an owner and operator, like Rivian, this isn’t a blocker for much of your workforce. So the opportunities in this space could be huge but only now being explored by companies who invested in BIM modeling (Rivian being one of them). I left this one as my last point because it’s less understood on where the value is generated but intuition to many in AEC suggests someone should spend some money here.

Below is a copy of the presentation I gave to successfully land the job of Automation Engineer at Rivian in the Design Technology team. We chose Automation Engineer as the title because the facilities organization didn't have a culture of tech roles and so the title needed to sound like it solved problems, not costs money.

First Steps

What Rivian's brand has done well is create energy around the product line and the lifestyle it supports. The people I met in my first weeks were SO jazzed about working and the future of the company. I found public Slack channels like #social-camping with photos from weekend trips and #emotional-well-being with advice on meditation. Each year employees were randomly selected to compete in an offroad rally where they used an R1T to compete against gas trucks. It was a contrast to my previous jobs and I was energized by it.

I supported a team of design technology managers who were in charge of helping define the way our facilities design and construction teams worked. They managed how we use Revit, AutoCAD, Autodesk Construction Cloud (ACC) and a couple of other platforms. It was my job to work closely with them to find opportunities for improvement with automation or custom tools. The whole org was lean, 100’s of active projects and only a handful of people managing; it was impressive (start-up vibes).

The first thing I setout to do was take my own advice from the creative exercise and set up data pipelines for our AEC SaaS platforms into a datalake. These platforms store the source of truth information for our active projects. RFIs, Issues, document sets, meeting minutes. It’s how our internal teams collaborate with the external AEC partners we hire to do the work. I learned that Rivian had a centralized data team. Their job was to run our company wide datalake (collect, store and govern access). This included ALL of the data streams like Workday, SAP, etc… When I met with their director to see what it would take to get our facilities platforms ingested he stopped me and asked how this data would be used to empower the business objectives. After creatively trying to frame facilities as key to the business he said, yes we can do it but we have dozens of more important projects in front of us that are REALLY key to the success of the business and it could be 2-3 quarters before they could start. He was a smart guy, not his first enterprise data director job, and I agreed that our team wasn’t a top priority and I put our requirements into the waiting list.

The company was modern. Imagine if you started a company today what the technology stack would look like. Amazon Web Services (AWS) was easily available to ANYONE who wanted it and you were assigned a Cloud Excellence Engineer who helped you design cloud applications. You had an assigned cyber security person who was in charge of approving any public facing endpoints but if you wanted to create tools and automations that ran on the internal-network, no approval needed. 

Since we had access to AWS I created our own datalake using AWS Athena (a pay per query database). You would be surprised how cheap and easy it is to create databases these days simply by having CSV files (as easy as Databricks and Snowflake and 1/10 the cost). The datalake got fresh data each morning and became the backbone of everything we built moving forward. But in the back of my head I knew that in the future we needed to move to our company wide datalake. This temporary solution would enable us to start whittling away at solutions to see what had true value. In addition when the enterprise data team was ready for us we would have the data pipeline built (a Python Jupiter Notebook) that they could use to speed up their work.

As I started meeting our design and construction managers I ran into a lot of tools built on what I would describe as post spreadsheet platforms:  Airtable and Smartsheet. These platforms empowered a new more tech-savvy workforce to solve problems that in the past were solved with custom development or spreadsheets. They store data like a database (forced into a schema, unlike Excel), provide multiple presentation layers and enable integrations through APIs or low-code. These platforms were the tools for many people at Rivian Facilities to get their day to day done when the product features fell short. You would see project tracking tools built to track what furniture vendors were on what service center for example. Or milestone calculators that would set task dates by working backwards from opening goals. What we noticed was there was a lot of manual syncing of data from the source of truth platforms into these tools. There was our first opportunity because these platforms had clean APIs to move data around and our datalake was already running.

Once the data was flowing we got such positive feedback from everyone. When someone makes a tool for themself that helps them do their job faster and better, and you come in and make it EVEN better; easy win. The ability for the digital workforce to solve their own problems without needing to always rely on external teams is such an amplifier because not only does it solve business needs, it gives employees a sense of agency and they feel creative. And this is where I was hit with some real world insight. Maybe instead of building your workforce tools you provide the support system for them to build their own tools because they understand the requirements better and have the right motivation to create solutions. Just like I had a Cloud Excellence person helping me deploy Athena here I was helping construction managers solve their real world problems.

Now let’s view this from the businesses perspective. There is a tension between letting your employees build bespoke tools vs forcing them to work in a more rigid environment. If you have worked in an office you have run across that one person's Excel spreadsheet that is SO complicated that NO ONE except for them understands how it all works; fingers crossed they don’t leave. We had some internal conversations about empowering people to build tools on top of Airtable and Smartsheet and there were two camps. One group thought we should stop for the reason above. My take was we should let people be creative because our department was still growing and trying to understand how it would operate within the broader company. Facilities construction wasn’t a core part of the business, it was a support. I 100% agree that if we were using these platforms to sell cars through the e-commerce platform that would be a big issue. My long term view was the tools that are built could be used as blueprints to what the business needed as it became profitable and operationalized.

The ability for a digital workforce to solve their own problems without needing to always rely on external teams is such an amplifier because not only does it solve business needs, it gives employees a sense of agency and they feel creative. 

Trying to Give BIM A Second Life

When I started my career in 2008 I was in San Francisco working for an engineering firm that specialized in hospital MEP system design. I was hired as a mechanical designer (engineer in training) but when I showed up for my first week on the job my manager told me they had just landed a 1 million square foot hospital project in Palo Alto, and it will be done in Revit and they have never used that software before. I remember my boss saying, "you're young, you can probably figure it out?" That's when my BIM journey began. I attended multiple events in the Autodesk Main Street office where people preached the future of building design and it was 3D, open (IFC) and going to revolutionize everything. I remember going to a presentation where they talked about how soon permitting departments will start accepting 3D models for review. Now when I write this in 2024 not much has changed since 2008 in the United States. Sure, Revit is a little more performant but the vision for BIM set forth in 2008 was akin to the vision for crypto in 2017. A lot of it didn't materialize at the scale the vision had been put forward. Today we all acknowledged that modeling buildings in 3D saves money because it better prepares us for construction but the information typically dies there.

How can we make progress in this space as an owner and operator? When I started talking to my team I was given some leads specific to our company use-cases. “Maybe we can help our operations people at our factory with maintenance tasks because the plant is HUGE and modeled in 3D?” or “We can extract quantities from our models to assist our teams that do procurement?” So we decided to start with baby steps. The question I had was, are people ready for 3D and for BIM (the information in 3D)? Most non-design jobs in facilities revolve around PDF files. Sadly because of the legal nature of construction none of that vision for 3D files being legal documents materialized in the last decade. Most people I interviewed said they used PDF files and oftentimes those files were outdated; they asked if we could fix that problem. And so there was our first real lead.

We store all of our design files in Autodesk Construction Cloud. We built a way (using APS) to access the files from a static URL that would always render the latest version and didn’t require a login. It used a similar pattern to Google Drive, where you can create a unique URL for a file that has a cryptographic signature where the URL is the password. If you have the URL you can see the file. If you wanted to guess the URL through brute force it would be near impossible. This app, which we called riv-viewer solved 3 major problems: 1. It always surfaced the latest version of a file, 2. It didn’t require the user to authenticate and 3. It rendered the file, 2D, or 3D in the Autodesk Platform Services Viewer which instantly enables viewing of the content. If the file was 3D your location and viewing direction was saved into the URL so if you wanted to reference something in a large model the URL took you there.

If you think about how we work digitally - we write emails and send Slack messages and if you can attach a URL in that communication stream that references BIM assets and the user clicks the URL and is instantly viewing, this is a feature. It’s very simple but very powerful. Riv-viewer became the most used app we deployed during my time at Rivian because it enabled the creation of communication that referenced large, or small, BIM or 2D data. We started seeing it used in Confluence and in SharePoint, in Email and Slack to enable our workforce to communicate about our facilities. The app enabled something the platform didn't allow. 

What I really took away from this project was the analytics. I would spend countless hours educating people about that riv-viewer and the app started to get used. We had 600+ live links circulating internal and with external partners. I added the ability to see what type of file URLs were being created - 90%+ were 2D PDF files but some were 3D Revit files. It was eye opening that outside of the design teams, no one wanted 3D. But there is hope because some of our vendors I could see used the 3D Revit files for what, I don't know but according to the analytics they spent hours in there. If I had the opportunity I would do some more user research in this space.

If you make accessing 3D as easy as opening a PDF file, viewership will grow and hopefully the value will follow.

Riv-Viewer was a platform. We later added modules that enabled extracting data from our Revit files that leveraged some of the company specific ways we authored 3D. For example we built a way to quickly count furniture because we had an internal team in charge of ordering. I wish I could say riv-viewer saved the company a lot of money but that’s hard to measure. All we see is usage and we hope the value follows. 

We All Became Data Engineers

During my time at Rivian we had many reorgs. If you look at my LinkedIn you will see 3 different roles because of 3 different reorgs. The last role I worked was new to me, as a data engineer for our facilities analytics team. This team was responsible for creating KPIs and analytics on how the organization spent money on facilities.

If you looked at the broader company there was a huge support system for data engineering. We had several groups whose sole purpose was to collect, store and govern data access. As I met my new team of analysts I started to learn about what it was like to operate in an organization that had this “platform” level of support because my question to them was - why do you need a dedicated data engineer? What I didn’t know about data analytics is it’s more akin to product development and what they needed from me was a way to iterate quickly.

If you’re a data analyst, first you have a specialty, maybe it’s finance, maybe it’s construction project management. Your job is to spend countless hours with the leaders who will be consuming your dashboards and reports to understand what has value; almost like a user experience designer spending time with end-users. The way to make progress is by creating something that could be used to structure a conversation around instead of endless collecting requirements. This is where they were struggling - our platform teams serviced everyone in the organization and turn-around times for a custom data pipeline was in the weeks, which slowed progress and killed creativity. In addition, what was even worse, because the only place our analysts had agency was in their dashboarding platform (Tableau) a lot of the rich business logic for data cleaning was being hard-coded into the dashboards, preventing reusing these clean data sets.

In addition to speed our analysts needed more control over the data architecture. A common request when doing financial reporting is over-time financial statements. Oftentimes the source of truth data sources were as-of-now numbers and so it’s the job of the analyst to stand up their own historical data. This was something they needed but couldn't do on their own and had little support from our platform team as these data-sets grow in size quickly and need maintenance to keep quries performant.

We decided to create a sandbox for our team where we could be creative. Where our analysts could see the whole data engineering process and even learn how to write data pipelines and manage databases in an environment outside of their dashboarding platforms. We envisioned this would speed-up our ability to create value in surfacing data driven tools and once we could see the value we would initiate the process of moving over to our official company data platform. This was important because the data platform was staffed 24/7 and designed to outlive our current organization structure (recall all the reorgs).

I wish we had Databricks when we started this project because it would have been our sandbox, but at the time it wasn’t an option so we used an existing AWS account because it was an approved approach internally. We settled on some core principles for our team:

• All pipelines will be written in Python using Jupyter Notebooks and it was everyone’s responsibility to learn how to write Python and work with our enterprise Gitlab. 

• If an analyst was too busy, the data engineer would write the pipeline or would provide code review and mentorship.

• When the analyst had a pipeline notebook ready it was the data engineers job to deploy it and if time, to educate the analyst how to deploy it.

• All data pipelines created new tables in our AWS databases (Athena) that were accessible to our whole team for reuse. We would focus on removing data cleaning logic from our dashboards.

• We only had two tiers of data storage: raw and transformed. Transformed tables could create new transformed tables.

Teaching a team of analysts how to use an AI coding assistant like Github Copilot enabled me to scale mentorship on the basic of data transformations with Python. We all quickly became data engineers together.

We ran this approach for about a year and I would say it was a success. We all learned so much from it and as a team learned so much from each other. I still stay in touch with some of that team and they have continued to use this sandbox to make progress prototyping and exploring facilities data sources.

When you look at the platforms available to companies today that enable a data driven business you see tools designed for data platform teams. We need to remember that the bulk of the challenge is not data engineering, it’s finding the value in the data (understand the right KPI for example). To find value you need nimble teams with nimble tools. We need platforms to enable analysts to self-serve in a way where their work can become the backbone of the business. In addition, we don't want to bog our enterprise data teams with requirements that may be half-baked and untested as their job is to develop and maintain the core data pipelines of the business not experiment.

Leveraging Strategic Partnerships

Every two weeks we would have our Autodesk account call. The customer success manager checks in and our consulting services folks update us on initiative progress. When I worked for Autodesk Consulting I used to be on the other side of the call but now I was on the customer side.

Customer success is a newer trend in account management. I have heard it explained that it’s the checks and balances for account sales teams. Customer Success Managers don’t get paid sales commissions and instead are measured by customer happiness metrics (annual survey results). As we moved from desktop license to the subscription models there is more incentive for account teams today to make sure customers are happy. Sales teams are measured on increased annual recurring revenue (not total) because it’s a sign that customers are finding more value in the products they are being sold.

At some point during one of our weekly calls the account team asked if they could present us with a vision deck. They had put together a presentation setting a vision for Rivian being showcased at Autodesk University. This conference is about inspiring people and they showed how our brand could be a good fit because both companies are trying to sell a future of sustainability. They asked to send this deck to our PR department and see if they have any interest; and so we did.

In 2023 Rivian was showcased at Autodesk University. I had the opportunity to present on what I wrote about in this article with my co-worker. The company had a showcase where the car was front and center in the vendor hall, with multiple other events that brought attention to the brand. This really taught me to view all of our vendor relationships in a more strategic light because after the event we all reflected on how mutual the outcome was for both companies.

Attached is our presentation that highlights some of the other projects we worked on during my time at the company:

Thoughts and Observations

After working at Rivian two years here are some takeaways I will remember when I start my next job: 

• You need to get your hands dirty. Large companies have support systems for most things but often you need to wait. If you can, just start and that enables learning. 

• The digital workforce is getting more tech savvy. If we give them the tools and support they need to improve their day-to-day using low-code and post-spreadsheet platforms, it could only help the business. We need to balance enabling creativity with business operation resiliency.

• Innovation questions the status quo. Digital workplace teams have become too strict when it comes to security. Facilities design and construction teams can benefit a lot from simplified security models because not only is most of the information benign but projects involve dozens of companies where the flow of information is the difference between on-time and on-budget.

• Giving BIM a second life is hard, but simplifying access and providing education could yield more effective teams. For design and build companies there are more opportunities. The ROI of digital twins for operators is still TBD.

• Data analytics is like product design. Analysts are on the front line of finding the value and they need tools and a support system that operates at the speed of their ability to make progress.

• AI Coding assistants like Github Copilot are a great way to scale your technical expertise as a team enabling more contributors as our tools become more technical and have coding interfaces, like in the data engineering example.

• Take advantage of your account teams. Sure, they are first and foremost trying to sell you services but there may be an opportunity somewhere in that relationship and if you’re open minded, you may find it.