Artikel zu AI + technischer Fortschritt

Must-read:
https://www.datasciencecentral.com/profiles/blogs/what-makes-a-successful-ai-company
http://www.bradfordcross.com/blog/2017/6/13/vertical-ai-startups-solving-industry-specific-problems-by-combining-ai-and-subject-matter-expertise

Nice-to-read:
https://www.datasciencecentral.com/profiles/blogs/comparing-the-four-major-ai-strategies
https://www.datasciencecentral.com/profiles/blogs/comparing-ai-strategies-systems-of-intelligence
https://www.datasciencecentral.com/profiles/blogs/comparing-ai-strategies-vertical-vs-horizontal
https://www.datasciencecentral.com/profiles/blogs/ai-strategies-incremental-and-fundamental-improvements

Full stack AI:
https://sifted.eu/articles/machine-learning-full-stack/

mydatastack/google-analytics-to-s3@2a6d5b0

Trigger alarm if no incoming data in S3 bucket within e.g. 24hrs. Should be defined in the parameters by customers.

• https://github.com/subkanthi/aws-glue-crawler/blob/master/glue-templates/glue-s3-crawler.yaml

Remove the lambda that does the configuration and change it to the cf configuration see above.

Classify ip to locations either during ETL or during validation / transformation phase.

Checkout performance penalty.

• https://www.upsolver.com/blog/small-file-problem-hdfs-s3
• https://www.upsolver.com/blog/aws-athena-performance-best-practices-performance-tuning-tips

• https://www.marcusstade.de/h2o-machine-learning-individuelle-vorhersage-fuer-die-next-best-activity-teil-1/
• https://www.marcusstade.de/h2o-machine-learning-stellenwert-der-microconversion-in-der-user-journey-teil-2/

Create a model for deduplication and proper tracking.
Inspiration? * https://help.amplitude.com/hc/en-us/articles/115003135607

/ga/source/event/year=XXXX/month=XX/day=XX/hour=XX/

• https://hevodata.com/platform
• https://www.confluent.io/blog/decoupling-systems-with-apache-kafka-schema-registry-and-avro/

1. how do I generate data, (trackers, dbs, scrapig websites, email, calls, other sources?)
2. how do I design business process that will create data for the algorithm that I want to build,
3. how to I handle the output,
4. how do I handle the process around the ml components -> how to integrate ml into existing products ~ https://www.youtube.com/watch?v=fB7nyxXaczY

There is too much focus on building the model, not much focus on how to integrate into existing products.

• https://transferwise.github.io/pipelinewise/
• https://hevodata.com/resources/videos
• https://www.stitchdata.com/

Hi
Can you write a brief description in readme file about your code? please explain how to execute the make file and their order also, please.

Based on the workshop:

• https://github.com/james-jory/segment-personalize-workshop

Data sent to the Import API is processed and sent to your destination through Stitch like data from any other integration.

Singer is a powerful way to write data integration jobs, called taps. Singer provides core functionality needed by applications whose goal it is to replicate data from a source to a destination on an incremental, scheduled basis. Common functionality provided by the protocol includes:

Persistent bookmarks for incremental replication
Authentication for common authentication schemes
Support for common data formats

What’s particularly critical, however, is that every Singer tap can be run within the Stitch platform. This is important because 80%+ of the cost associated with a data integration is in the maintenance phase. With your tap deployed on Stitch, you won’t have to worry about:

hosting a server where jobs are run
scheduling jobs
viewing log output of jobs
building notification systems to let you know if there are run failures

• https://blog.fishtownanalytics.com/why-we-build-singer-taps-for-our-clients/
• https://www.stitchdata.com/platform/extensibility/
• https://www.stitchdata.com/docs/integrations/import-api/
• https://www.singer.io/
• https://github.com/singer-io/getting-started
• https://transferwise.github.io/pipelinewise/index.html

• https://github.com/aws-samples/aws-etl-orchestrator
• http://www.lib4dev.in/info/aws-samples/aws-etl-orchestrator/132656351

• Blinkist implementation see slide 29 https://www.slideshare.net/SebastianSchleicher/tracking-and-business-intelligence
• https://medium.com/buildit/a-b-testing-on-aws-cloudfront-with-lambda-edge-a22dd82e9d12
• https://poplindata.com/privacy/itp-21/
• https://omr.ruhr/google-analytics-itp-2-1-prevention-http-set-cookie-snippet-182092779d40
• https://www.simoahava.com/analytics/itp-2-1-and-web-analytics/#server-side-analytics

Currenly the redeployment happens in manual mode.

This exec tearsheet is our most important dashboard. Each dashboard included the metrics that the teams were responsible for in the exec tearsheet, and also included other supporting metrics. The metrics that we measure exist in relation to our engagement funnels. You can think of three funnels at 500px: 1) visitor -> signup -> daily active -> daily engaged -> paid subscriber 2) visitor -> signup -> photo upload -> photo submit to marketplace -> photo sold on marketplace 3) visitor -> signup -> purchase photo from marketplace

Each team owns different parts of this funnel for different products:1) The marketing teams own (page views) and the top and bottom (revenue) of this funnel. 2)The product teams has less of an emphasis on top of funnel metrics. 3)The development teams (web and mobile), want to see the entire funnel with respect to their own products.

Source: https://medium.com/@samson_hu/building-analytics-at-500px-92e9a7005c83

Parse user_agent to get and spread into a complex data structure for further analysis.

• https://tealium.com/products/audiencestream/

• Find partner agencies, companies with a high leverage (labor + capital) that are going to promote the solution to their customers (e.g. pandata, leroi, trakken)

• Mentioned by Matthias see case w/ Tobias.

• Agencies that have big customers in their portfolio, customers to which you won't have access as a 1-men-show.

The bots should be filtered out during validation or firehose transformation.