We are looking for a Machine learning scientist to join our Science team and help us scale our ambitious Natural language processing (NLP) and ML (Vision, Speech) efforts to build the best conversational Artificial intelligence.
So far, each time we’ve incorporated new NLP models into the product, we’ve seen significant improvements to both the conversation quality, and the long-term user retention. In particular, our coreference resolution module (NeuralCoref) helped us understand the context in quick exchanges of short text messages measurably better. Our emotion classifier, called TorchMoji (trained on 1.6 billion tweets!) has let us create stronger empathy and more emotional connection to our users.
Your task will be to make our NLP stack and models 10x smarter. You’ll be able to train on our large scale conversation dataset (+100M messages), and you’ll see the output of your work in production, used by real users, in almost real time (we ship NLP models every week).
Along the way, you’ll also publish open source implementations and contribute to the state-of-the-art in academic research in Machine learning. In 2018, our team already got papers accepted in several top conferences in AI/ML happening this year: ACL (current top conference in NLP) and ICLR (founded in 2013 by Yoshua Bengio & Yann Lecun) and we are on track to be featured in several other top AI/ML venues this year.
What You’ll Do
- Design and implement cutting-edge NLP models.
- Working with our engineering and product team, deploy those models in production to our conversation stack, and monitor and analyse their results.
- Along the way, publish some kick-ass research papers to top-tier conferences.
Who You Are
- MSc. or PhD in Computer science, engineering or a related field (Math, Physics, Statistics, …).
- Vision alignment: you believe everyone will talk to an AI everyday in 5 years and you are passionate to work hard to achieve this vision.
- Proficient in Python and at least one ML framework: PyTorch or Tensorflow.
- Familiarity with NLP “deep-learning style”: RNN, CNN, attention-based models, embeddings (the content of Richard Socher’s Stanford class for instance)
- Optional but appreciated:
- Familiar with “classical” NLP tasks and tools: syntactic and semantic parsing, semantic relations extraction (openIE), co-reference resolution
- Knowledge of C/C++ and Carnegie Mellon’s DyNet (a C++ neural net framework) would be appreciated
We want to develop a strong culture of ownership where all team members are top-level builders and doers (no long meetings here!). We’ve done other startups in the past, so we want to always find the intersection between what makes you most excited, and what makes sense for the company.
This is a full time salaried position, including competitive salaries, stock options and health insurance. Location is in Brooklyn, New York.
If not based in NY we can help you relocate.
We are on a mission to build the world’s most used conversational artificial intelligence.
Our users already exchange one million messages per day with their AIs and the usage is growing fast. We operate on messaging platforms, inside our native apps, and in voice-first products. We are lucky to have awesome investors (first investors at Instagram and Snapchat) who support us fully in our vision.