Tell us about what you do at Dune 

My role at Dune is Staff Software Engineer, Machine Learning and AI, which means I have a difficult time explaining to my parents what I do for a living. 

I work on features that use AI and machine learning to improve the creator and consumer flows. That can range from implementing tried and true algorithms like PageRank to bubble up useful tables to the top of our Data Explorer and building LLM based SQL debuggers to fix persnickety syntax errors.  

How did you get into Engineering and more specifically AI? 

Right time and right place. I moved to San Francisco after college and worked as an environmental engineer doing fluid dynamic modeling work on stream and wetland restoration projects. I was flat broke and rented half of a converted garage with an electrical engineer. I thought about moving away, but then I fell in love and knew I had to figure out a way to make more money. So I packed up all my math and programming skills and said sayonara to saving the salmon and got a job as a data scientist at a large gaming company. It was a great time to get a job in San Francisco and my future husband quizzed me on SQL queries in Dolores Park in preparation for the interviews. 

What does a day at Dune look like for you? 

I work on projects from ideation to deployment to maintenance, I balance my day between forward-planning, implementation, and bug squashing. AI and Machine Learning work can be very speculative and I work on loose mandates like “make it easy for creators to do X” or make the feed more “trustworthy” which means it’s up to me to ideate on how machine learning and AI can help achieve those goals.  If I am early in the stages of a project, I spend my time on a literature review to gather ideas from academics or other companies facing similar problems. I love this type of work because you can get so many good ideas skimming arXiv articles.

Once I have a good idea,  I build a “benchmarking” dataset. As a machine learning engineer, I love a good algorithm but the first step is always thinking about a good way to benchmark because you might find that a heuristic will get you 95% of the way to your goal. This can actually be a fairly creative exercise because real problems don’t come with Kaggle CSV downloads. When I have my benchmarking dataset, I start testing different algorithms and heuristics, once I’ve got a good solution, I move onto implementation and writing production code. Once the feature is out and tested, I monitor it and squash any bugs that arise. 

What has been your favorite Dune moment and why? 

Probably playing Hammerschlagen, an Austrian drinking game, at one of our company offsites after a day of skiing with the team. Since we are a remote company, it’s a lot of fun to get together and see each other in person. Especially, if there is a little friendly competition! 

What are you most proud of? 

I’m most proud of Spellbook, our open source data pipelining project. This is the least AI and Machine Learning product I’ve worked on, but I love that it’s a place for the broader data nerd community to work together. I worked hard to launch that project and I’m happy that it’s still going and our biggest challenges with it are that it’s too big!

What three words would you use to describe Dune?

Fun, Fierce, and Family Oriented

What do you love most about Dune? 

I love that I get to work on challenging problems and still get lots of time to be a mom. Many women I know take a big step back in their careers when they have children but I feel so supported by my manager and the company. 

What are you listening to on repeat right now?

Hostile Government Takeover 

Tell us a fun fact about yourself

My fun fact is I taught myself (some basic) piano during my most recent maternity leave and wrote a blog on machine learning and biology applications. 

Connect with Meghan:

LinkedIn
Medium