TRASH is a new video tech startup in Brooklyn. We’re using AI to build predictive editing for video in fun, ethical, and experimental ways. We think a lot about machine assisted creativity, the rise and future of mobile video, and the intersection of art and technology. We also think about what makes a good house party “good”, and if the principles and outcomes of this can be applied to a new corner of the internet :)
If you get excited by the idea of having a ton of autonomy and leadership at an early stage startup, learning and making alongside two repeat founders with deep domain experience, and helping shape an inclusive and diverse product and company culture, this is for you. We’re early stage (pre-seed funding, pre-launch) and are ready to build out the data model for our core tech – which is where you come in!
We’re looking for an interdisciplinary data scientist with a strong visual and ML background who has a passion for wrangling nuanced data sets, solving difficult creative problems, and isn’t afraid to bleed from the cutting edge of what is *just* technically feasible. You must be able to code (though not necessarily deploy to production), as well as use your ontology engineering / information science skills to do topic modeling and classification on video data.
You would be working with us to…
- Build a set of taxonomies, datasets, and trained models that can map the relationships between video data and video aesthetics (eg. sentiment, saliency, scene, pop culture references, visual trends)
- Help develop specialized networks, sequence-to-sequence models, and use transfer learning techniques for broader objectives than naive classification
- Define and measure the effectiveness of these classifiers, both with automated and manual testing techniques
- Advise on database solutions for housing the data you will be mapping and classifying through helping the team think critically about video taxonomy
- Use your data driven insights to help drive product design decisions, as well as build light weight internal tools / dashboards to share your insights with the team
- Contribute to Synopsis, the open source platform for video metadata analysis that TRASH is built on
- Help shape the future of TRASH!
- Solid grounding in statistics, probability, data modelling, ML algorithms
- Understanding of software development techniques and toolkits
- Familiarity with programing languages like Python, Ruby
- Working knowledge of ML/AI frameworks like Tensorflow, Torch, Caffe
- Experience with deep learning/neural nets in production and/or at scale
- Experience building working with classification and inference
- Understanding of bias and the ethics of data set collection
- Ability to work with messy, unlabeled data sets. A pragmatic, “hacker” mentality
- Experience deploying ML models
- Experience with consumer-facing products and putting the user first in all your decisions
- Experience working for an early-stage startup and wearing many hats
- You are a creative human who cares about design, style, fashion, music, photography or video, and maybe even dabbles in the subjects.
This position is freelance/PT in New York with potential for FT (early stage equity + salary)
If this sounds like an Exciting Thing to you, please email us to say hello and include a copy of your resume to email@example.com