PAPIs.io '15 is the 2nd International Conference on Predictive APIs and Apps, taking place right before KDD. PAPIs is the premier forum for the presentation of new machine learning APIs, techniques, architectures and tools to build predictive applications. It brings together practitioners from industry, government and academia to present new developments, identify new needs and trends, and discuss the challenges of building real-world predictive applications.

Tutorials-----------------------------------------------------------------------------------------

Tutorials are long presentations that focus on teaching valuable skills (e.g. "how to do this in your company"). They shouldn't be product-oriented or rely on commercial software. Tutorials should target those who are not experts in Machine Learning (ML) or APIs. Here are a few ideas of topics to inspire you (this list is by no means exhaustive and these are just ideas):
• feature engineering and data transformation for predictive apps
• how to measure performance for model selection and live monitoring
• principles and real-world applications of bayesian ML / deep learning / etc.
• how to turn your ML back-end into a predictive API
• how to use churn detection in your company
• etc.

Research presentations-----------------------------------------------------------------------------

For the first time this year, we are introducing a research track for people who create predictive APIs and who develop Machine-Learning-as-a-Service (MLaaS) systems. Topics include (but are not limited to): 
• Software engineering: design patterns and best practises
• Distributed systems: scaling out services and APIs
• Machine Learning / Data Science automation
• Interoperability between services / APIs / tools
• etc.

Main conference presentations-----------------------------------------------------------------------

The typical format is 20' presentation + 5' questions on predictive use cases, stories and lessons learnt while building real-world Predictive Applications and Predictive APIs. Topics of discussion include (but are not limited to): techniques, architectures, tools, challenges, needs, trends, impact, etc. Keep in mind that because of the single track the audience will be mixed (see this blog post about attendees): API guys, Data Scientists, Developers, and non-technical types. It’s ok to address a subset of the audience at some point in your presentation, just make it clear for the audience and do it reasonably. If your target audience consists of experts, you should consider a research presentation instead (see above).
Preparing your presentation abstract and outline: We believe that great presentations should be practical and focused. Instead of presenting something general or showing a portfolio of things you and your company have done, it’s better to show one specific, unique thing in enough detail. Presentations will be 20’ long which is a short time but allows for dynamic talks that keep the audience engaged. To help you prepare your proposal, we’ve listed below some things we're interested in seeing:
• problems / challenges you've overcome or are currently facing
• stories of what you did, how, and lessons learnt while building your predictive API or app (do’s and dont’s)
• tools / libraries / APIs / services you used, why you chose them, and your critical opinion on them
• how you architected what you built and why
• the data you're dealing with (which? how much?)
• the non Data Science / Machine Learning aspects: infrastructure, software engineering, User Interface / Experience, etc.
• how people are using what you’ve built and how it’s creating value for them
• what makes it novel
• how you measure performance (offline and live) and which results you got
• what you plan to do as future work
• hands-on stuff (screenshots and live demos)
Can I feature a particular tool or product? Yes, you can if your presentation is hands-on, centered around real-world use case(s) and includes practical information. Do not just demonstrate your raw toolbox but highlight unique features and demonstrate practical integration(s) into customer/end-user application(s). Make sure you do not turn your presentation into a sales pitch as our audience dislikes them and our objective is to ensure that PAPIs does not turn into a platform for pitching products / services / companies. It is best to submit your proposal early to get our feedback and adjust before the deadline if necessary.