Earlier this year, Grant Ingersoll and I kicked off a 4-week co:rise course on Search with Machine Learning. We covered search fundamentals, and then went on to teach ranking (both hand-tuned and LTR), content understanding, and query understanding. Students worked with a real (if slightly dated) ecommerce product catalog, developing and improving a search application as they learned new techniques. Our goal was that our students would take away not just theoretical knowledge, but practical skills they could apply in their day-to-day work — whether they were already search practitioners or looking to enter the industry.
Teaching 127 students was a daunting, humbling, but ultimately very gratifying experience. Both of us had taught before, but neither of us had ever attempted to teach this quantity and depth of material at this scale. Moreover, it was a truly global class, with students signing on from the Americas, Europe, Asia, and Australia.
We’re grateful to the folks at co:rise not only for giving us this opportunity to share our knowledge, but also for supporting us and our students every step of the way. And we were overwhelmed by the level with which students not only engaged with the material, but also supported one another. We know that the material presented many conceptual and technical challenges, and we also had our share of rough edges in this inaugural class. We know that many of our students had to do a lot more work than they had signed up for, and we hope that the knowledge and skills they took away justify that investment.
Here’s a sample of what our students shared after completing the course:
I had a blast! 4 weeks of intense immersion, with great instructors, and a very open and sharing community I now have new tools in my toolbox. And new additions to my professional network to ask when the need arises.
— Fredrik Olsson, Head of Data Science and Product Owner @ Gavagai
Excellent, it is more than a course, it is an immersive experience.
— Jesse Peixoto, Search Relevance Developer @ ADEO
Thanks to Corise and all the people who made this happen, I can not think of a better way to learn than how the course is organized. Learning from experts in the field, well-organized weekly projects, and a great community of learners are the highlight of the course for me. Grant and Daniel have gone above and beyond to simplify all the concepts and clarify every point of the course material. I can genuinely feel a sense of achievement after finishing the course. I truly appreciate all your work : )
— Mohammadreza Alagheband, Search Engineer @ Tidal
This course makes it real! Not just slides, but real use cases, with real problems. Combined with access to experts, this course was SUCH A BLAST!
Thomas Queste, Technical Lead @ ADEO
We’re fiercely proud of this inaugural class, but it’s clear that there’s room to improve the course. We received lots of constructive feedback from students, leading us to prioritize the following:
- Reducing busy work. We tried to minimize busy work in the projects, but we still saw a lot of unnecessary pain in setting up the ingestion and LTR pipelines. Experience is the best teacher, and we hope those who felt this pain will take solace in knowing that future students will benefit from their experience.
- Covering dense neural retrieval. We taught students how to use embeddings for content and query understanding, but we only touched on vector-based methods for indexing and retrieving content. We know that many students were hoping for more, as evidenced by the popularity of a community talk that Dmitry Kan presented on the topic.
- Supporting beginners. We pitched the course as having no prerequisites other than basic Python programming. But the course attracted students with a very wide range of backgrounds, and as a result we struggled to establish a pace that worked for everyone. In retrospect, it’s clear that beginners would have benefited from more support.
In order to address the last two issues, we decided to split the search course into two courses:
- Search Fundamentals: a 2-week introductory course intended for people who have no previous experience working with search.
- Search with Machine Learning: a 4-week course that assumes a background (e.g. basics of indexing, ranking and aggregations) in Search Fundamentals and will cover advanced topics like dense neural retrieval. This course requires certification in Search Fundamentals or equivalent experience as a prerequisite.
We hope this change improves the experience for both beginning and advanced students, and we’re excited to continue cultivating the incredible community we’ve just started to grow.