Smart Code: Building and deploying AI business projects

How to do DevOps, efficiency calculations, ROI predictions, algorithm choices, and software stack re-engineering for AI-centred projectsAbout This Book* Practical approach to AI* Providing context, ie web or cloud, for AI* Teaching AI project execution* Relegate purely technical considerations of AI algorithms to their proper placeWho This Book Is ForSmart Code can be read and understood by programmers and students without practical AI experience, since it is meant as an introduction to AI in practice. By the same token, some AI technical reading or AI algorithm implementation experience would be most welcome, whether as part of an undergraduate or a business software project. Further, knowledge and experience of Python programming is required to understand the code in this book.What You Will Learn* How to fold AI algorithms into existing software stacks* Building AI software that is relevant to business* Choosing AI algorithms according to the nature of the data they have to process* Adapt pre-existing systems to the demands of online AI software* Plan cloud infrastructure for AI applications* Turn new data into a gold mine for the discovery of new trendsIn DetailArtificial Intelligence begins with expertise in algorithms, but its practice does not end there. The effectiveness of AI algorithms is often determined by factors unrelated to their sophistication. AI applications sometimes have to be folded into existing software stacks. Pre-existing analysis software has to be either taken into account or removed. There are no hard and fast rules to accommodate these cases, but there are approaches and techniques which Dr. Eckroth explains in this book.Smart Code starts out with a project that attempts to resolve a fairly common conundrum, namely how to reconcile the existence of AI software on the cloud with the needs of a business that needs quick results and snappy response times.Another, very common scenario is the existence of user feedback: it is produced in an almost batch-like fashion, unpredictably, sometimes idiosyncratically, yet analysis is supposed to yield useful results. This chapter walks through a scenario telling us how to produce results reliably, without assuming an overly predictable, smooth data stream.The next chapter talks about the classic case of a recommendation engine in a business environment. It is easy to take recommendation engines for granted, but the fact that recommendations are produced by unpredictable users makes running a complete AI stack somewhat tricky.Deep Learning algorithms have been prominent in image analysis and pattern recognition. There are systems and performance considerations that Dr. Eckroth will be talking about, and he will also provide a case study about trend detection on your own website.Finally, the question how to generate replies to user queries is common, but difficult to answer just by referring to the appropriate algorithm. Dr. Eckroth dives into questions of relevance and choice of techniques before the final chapter on AI system trends influencing the future.