Twitter is what's happening in the world and what people are talking about right now. From breaking news and entertainment to sports, politics, and everyday interests, see every side of the story. Join the open conversation, and collaborate with creative and curious people across the globe.

The whole world is watching Twitter. You don't go a day without hearing about Twitter, how it's used as the fastest way to send a message to the world in an instant, how it carries some of the most important commentary and conversations, how it mobilizes people into action. That's powerful, it's valuable, it's fundamental

Twitter is seeking a Senior Machine Learning Engineer for Birdwatch, our pilot program in a new crowdsourced/participatory approach to reducing misleading information. Birdwatch pushes the state-of-the-art in approaching misleading information on the Internet and we employ a deeply experimental, fast-moving, and iterative approach to find product solutions that work for customers.

We are looking for someone who can design, develop, and implement the core algorithms and machine learning models that find and label the best Birdwatch notes from the crowd, ensuring that Birdwatch’s voting and reward systems are resilient to adversaries and bad actors. You’ll work as part of a cross-functional team including machine learning, engineering, data science, research, design, product, and even academic experts outside the company who study the space. 

What You’ll Do:

As Birdwatch grows beyond its pilot phase, you will play a critical role in designing and building the core algorithms and machine learning models that power Birdwatch in real-time, including:

• Compute contributor helpfulness scores that are resistant to adversaries and bad actors, e.g. using iterative graph propagation algorithms in a similar style as PageRank, or explicit coordinated manipulation/spam detection

• Detecting rater similarity/diversity and polarization, using techniques such as matrix factorization to learn user embeddings or similarities

• Determining which contributors to ask for ratings from in real time, and determining overall labels from the crowd, in order to display Birdwatch notes on potentially misleading Tweets quickly

• Working closely with human-in-the-loop teams to design the best ways to evaluate our algorithms and ensure data quality is high 

• In the spirit of Birdwatch’s transparency, open-source as much of our core algorithmic code as possible

• Collaborate with product, design, research and eng to shape the product experience of Birdwatch (e.g. what questions do we ask in voting, to whom, do we add friction to parts of the process to reduce potential manipulation, etc) to help achieve the above goals.

• Collaborate with external researchers as well as other Twitter teammates who have expertise in the area to design our approach

• BS, MS, or PhD in Computer Science, Machine Learning, or a related technical field, or equivalent experience

• Very strong knowledge of machine learning and statistics

• 3+ years of experience in ML software engineering, including designing/training models and building ML data pipelines

• Proficiency with Python and a modern machine learning framework e.g. Tensorflow or PyTorch

Bonus qualifications (but not necessary):

• Deep expertise in relevant domains, e.g. graph algorithms, matrix factorization, using human-in-the-loop data, bad actor/manipulation modeling, game theory, interpretable and fair ML, active learning, etc.

• Proficient in SQL and one or more of the following: Scala, C++, or Java

• Experience with large-scale production environments and ML data pipelines, and with specific technologies you are likely to use at Twitter building them e.g. Kubernetes, Kubeflow Pipelines, TFX, Dataflow/Beam, BigQuery, Airflow

Who You Are:

• You thrive in the uncertain environments, e.g. you are able to dive in and use product intuition and data science when necessary to understand your data and decide what objective functions or model types it makes sense to use to meet our product goals

• You are experienced with machine learning best practices, but know when machine learning isn’t the best tool for the job and are comfortable reaching for other types or algorithmic or statistical approaches instead in those situations

• Enioy working in an ambiguous space with no known product solutions, and trailblazing to build novel solutions that work

• Passion for the problem of misleading information & creating a better informed world

• Enjoy a rapid iterative approach / the 0-to-1 experience

• Ability to move fast and get things done, and help the broader team do so (to enable rapid iteration on the product)

• Comfortable (maybe even excited) to work with a distributed team

All your information will be kept confidential according to EEO guidelines.

We are committed to an inclusive and diverse Twitter. Twitter is an equal opportunity employer. We do not discriminate based on race, ethnicity, color, ancestry, national origin, religion, sex, sexual orientation, gender identity, age, disability, veteran, genetic information, marital status or any other legally protected status.

San Francisco applicants: Pursuant to the San Francisco Fair Chance Ordinance, we will consider for employment qualified applicants with arrest and conviction records