Shipping AI That Matters: My Path from First Dashboard to 2 Billion Users

Hi there! I'm Cassandra, a senior product manager, passionate about empowering teams to grow, staying curious, and delivering more than they thought possible. Below is my journey shipping AI/ML products from my first dashboard to my first launch to 2 Billion users.

Data Analytics @ Texas A&M University & Early Career

My love of data science began in college as I studied finance. I fell in love with corporate valuation and the use of calculus for finance use cases. I was then fortunate enough to secure roles in my early career as a business analyst and marketing analyst, where I had the opportunity to analyze business and marketing campaign performance, build SQL automations, no-code apps, and financial models to reduce monetary waste and secure $32M in funding.

Dashboards, ML Evaluation, & Ops Tools @ Meta

At Meta, I joined an operations team where I built a SQL-based dashboard that showed the performance of the operations team in terms of the accuracy of misinformation content review.

I stayed curious, had coffee chats with Shengbo Guo, a senior ML computer scientist, and learned how supervised models work and how they're evaluated.

Curious about this space, I took on side projects to help the news feed integrity infrastructure team implement models into the news feed ranking algorithm. It was this curiosity that allowed me to join their team as a project manager.

While working with this team, I led the fine-tuning of an ensemble machine learning model to reduce misinformation on Facebook by 30% for over 2 billion users.

Additionally, I led the development of another suite of tools that enabled operations teams to review, label, and analyze viral content to understand which elements of news feed ranking contributed to its distribution, and increase the speed of ops investigations from 7 days to 24 hours. 

Developer Experience Tools @ VMware

As a senior product manager in VMware's business services organization, I was embedded with the U.S. Army, tasked with teaching them about lean software development.

One of my favorite projects was standing up a developer experience team. It all started with a pain point: my team was ready to ship our MVP and sent our product for review with the security team (security is essential for the Army, as you can imagine).

The review was lagging, taking a lot of time, and had few updates. Curious about the issue, I organized the security team and my engineering lead, Jason McKee, and we mapped out the user journey of the average app team as they work through the security review process.

Together, we identified pain points and committed to developing better solutions. I put together a pitch for the ASWF leadership to make the case for establishing a team to address developer experience pain points and develop solutions.

I guided my team through user research conversations with the 18 or so app teams in the organization, developed a roadmap, and launched a CI/CD product, decreasing security review time from 90 days to 48 hours, and saving the US Army, Navy, and Air Force an estimated $4.8M in the first 6 months of launch. Not only did this save a massive amount of time for the organization, but it also set a precedent for user-centered design in infrastructure spaces. See this write-up on how to lead infra teams with user-centered design principles.

Excited to foster a data science culture for the Army, my colleague Amanda Whitehead and I decided to write a framework on how to maintain a balanced team with data science teammates. And more specifically, how to develop AI products from a balanced team perspective. We decided to give some webinars on how to develop AI products (with Jordi Ensign), and how to manage AI products.

Designing AI Products @ MITx

Curious to continue learning, I participated in the MIT course on designing and building AI products, where I gained valuable insights. I then took Python, computer science, and calculus courses at my local community college in preparation for a Master's of Data Science program.

Building AI Prototypes @ Antler VC & Buckled.io

As a founder, I developed a SaaS solution for the automotive services industry with a 5-year, $3 billion valuation roadmap. I joined the Antler VC accelerator, refined the pain point, and reimagined the MVP, which culminated in a hackathon within the first month of the accelerator. We used Cursor, Eleven Labs, and Lovable to build a prototype that would give you an estimate of how much a given repair would cost. We hustled for 48 hours and won the hackathon.

I then built a RAG pipeline that utilizes a database of car service manuals, the ChatGPT API, and delivers a personalized quote for your specific car and service needs.