Launch HN: Turing College (YC W21) – Online data science school
When the three of us entered university, we were taken back by the outdated teaching methods. We still smile when we remember learning Excel via a whiteboard! While studying, we were also running an IT and education consulting business that had an accounting return rate (ARR) of $0.6M and was expanding fast. This quickly taught us that the way developers are educated is not aligned with the hiring and onboarding processes of the tech companies looking to hire them.
We saw this issue from both sides of the hiring process. As students, we were learning subjects that didn’t prepare us practically to deliver results for companies from day one. As employers, we were frustrated when hiring students based only on their educational credentials, as these weren’t a good guide to future performance. So, we decided to organize a non-profit data analysis bootcamp, where the curriculum was supplemented with hiring partners' projects. First batches were oversubscribed and we were nudged to build a school, which would create specialized data science courses.
Our programs are self-paced, so we’re not a bootcamp in the sense of forcing more and more information on people each day, whether or not they have digested the previous material. Completing a course with us usually takes 9-12 months, but students can progress as fast as they like, and some experienced software engineers have completed 1,000 hours of coursework in 6 months. Conversely, students who are transitioning to data science from other fields, and lack fundamentals in maths or statistics, can go slowly and build solid foundations in these areas.
Students choose between several data science specialisations, including data analysis, requiring a solid understanding of statistics and mathematics and excellent data wrangling skills so that data analysts feel comfortable importing, cleaning, and manipulating data; and machine learning engineering, focused on building machine learning models that solve business challenges. Our curriculums are co-created with tech companies who we partner with, who tell us specifically what they are looking for in new hires. Since we started 6 months ago, we have had 17 companies contribute to our learning concept, including Moody’s and NordVPN. We teach current tech stacks and use specific problems companies have worked on as the basis for projects that students work to solve. These later turn into project portfolios that help them get hired.
Each student also gets regular industry professionals guidance from our staff and hired Senior Team Leads, working professionals in the data science field. They perform 1-on-1s, standups, do mock-up interviews, and more. These professionals are paid consultants who joined us from Waymo, Unity, and more. One student writes: “I studied in university, and at other coding schools, but Turing College is just something totally different. The best part is the ratio of personal tutoring hours we get - it is 10x more than in the places I have tried before!” We use standups and 1-on-1s with senior leads, and students get a weekly minimum of 3 hours of personal consultation with their leading peers and/or senior staff.
Students also get feedback, motivation and encouragement from their classmates. We have a diverse community, including fresh graduates in STEM subjects looking to specialise, right through to software engineers who want to enrich their data knowledge. This diversity enables mutual support. Those with backgrounds in maths and statistics can help those with pure coding background, and those with experience in business can support with soft skills. It’s a collaborative, community-oriented approach that we support and encourage through reg...
62 comments
[ 2.9 ms ] story [ 142 ms ] threadEdit: I'm also really glad you're not selling ISAs. To me the direct connection/ownership of the debt is what really aligns the parties, not the ISA itself. Once it's sold off to a third party debt collection agency of course there is still some alignment but I think the connection gets fuzzier.
My piece of a feedback. "Good data scientists are in huge demand" -- that's what everyone says. But I believe "good data scientists" are usually PhDs with a few years of experience.
I don't want to say your program isn't good enough. I think if you can educate someone to have an entry Data Science job in 6 months then it's a great success. By "entry DS job" I mean a real DS job: not Excel munging.
Your 6 months course seems like a bit of everything: which is fine because it's an entry course. But it's not much different from any other entry-level course. What differs you?
This is the same argument that says only people with CS degrees can be good software engineers. That is provably wrong because there are good software engineers who don't have CS degrees. Good data scientists are just people who can do data science well. That's the only valid measure. Having a PhD and some years of experience increases the likelihood of that being the case but having those things doesn't automatically mean someone is good, and not having them doesn't automatically mean someone is bad.
This would require learning calculus, then statistics & probability, then linear regressions & generalized linear models, then machine learning(NLP, machine vision and/or AI). I’m not sure how someone can learn all of that in one year let alone 6 months with no prior experience in the field.
I can sort of see it being done but it would have to be very superficial knowledge which is sort of useless when your trying to recreate some paper that uses advanced stats and math. You could maybe recreate the paper but you would essentially being shooting in the dark and you wouldn’t really know the limitations of the model in any meaningful way.
Im curious to know how they will address this?
I don’t think you need a PhD in statistics or machine learning to be a good data scientist, but I’m just having a hard time believing all the skills can be learned in less than one year without any experience.
The thing with data science that it is a pretty new field, and data scientist tasks differ from company to company. Partnerships with companies help us understand the maturity of data science in every company and prepare students accordingly.
What do you mean by this exactly? You are school, right? It’s not possible you are not teaching them right? Should a person not have the right to fail the course from start to finish?
The reason we are doing this is that at a certain point knowledge starts to "fade away" and even if you are in the middle of a course, you might start forgetting things you learned at the start of it. That said, self-paced learning is unfortunately not something that works for everybody. Our students get to try it out in our admissions process & demo month[1] to cancel free of charge before they commit.
[1] https://www.turingcollege.com/faq/what-is-the-demo-period-an...
For the purpose of trying to determine what is or is not fair I don't think it matters whose fault this is (a failure of the school's pedagogy or of the student's ability to keep up) because I think allowing the failure state to continue is fundamentally unfair.
1) Our program is self-paced, while almost all bootcamps have a strict learning schedule. We ensure accountability through daily meetups, 1on1, and project assessments. Turing College suits people, who work full-time and want to upskill part-time by flexible schedule. As we have platform, which organizes all the learning activities for students, project-assessments happen from Monday-Sunday from early morning to late evening without our operational staff involvement.
2) Students spend 1/3 of their learning time on hiring partners' projects. Companies create projects that reflect their business problem and tech stack. Thus, our students are more ready to deliver from day one than other bootcamp alumni.
3) Students learn everything through practice. I know this could sound cliche as many bootcamps claim to do that, but in reality, it is just happening on paper. We have a custom platform that organizes curriculum into data science projects, which each of them should be assessed by a minimum of three people. To finish Turing College, students need to complete 20 projects that should be peer-reviewed by Senior Data Scientists and peers. Through this process, students get a lot of feedback about their learning and overall performance. They also need to assess at least 20 projects to finish the course - by this, they are pushed to learn to examine other's work, which is crucial in every technical position.
4) Students have personal development programs to improve their habits, communication skills, critical thinking, etc. Many employers' problems are related to the human factor, so we focus on shaping essential soft-skills parts.
Disruption happens by innovative, open minded people. But boardrooms of most companies and admission committees of unis are filled with legacy people most of whom are there because their parents put them there. For academia, there are only people who took a traditional path for education. And they only value who are trying to get there by those paths only.
First, higher education sector and industries need to validate the the disruption for that disruption to be really meaningful.
1. The total cost (5k direct[1], 9k ISA) is significantly cheaper than Lambda: do you see that as a reflection on the cost sensitivity of the European market or is there a reason your model allows for lower-cost execution?
2. The ISA model gives the school a vested interest in the career outcomes of students, but Lambda has highlighted that ISAs can become a debt product that is resold: do you see that model (of reselling ISAs) as part of Turing College's future?
3. Lambda has had problems with new courses being unable to deliver on the promised quality: how do you see the next few years of growth for Turing, do you expect to introduce new courses based on demand or do you expect to build out new curriculums (and test with whole cohorts) before introducing them?
4. What's the relationship between supervisors and Turing? The team page notes that these people are part-time with Turing: how do you ensure that they're able to deliver the valuable mentorship required by students? Are they volunteering? Paid per hour? How does their mentorship accommodate students that require more support than average? Does their commitment to Turing come before their work, with their employers supporting?
Very promising proposition, team etc: very interested to hear answers to the above to better understand how you are approaching the more challenging aspects.
[1] I'm an employed Software Engineer but at that price it's very tempting to enroll to build out my data science skills.
1. You are correct to say that our pricing reflects the European market and is naturally different from schools in other markets. In addition to that, we are a self-paced school and have no actual classes happening (work-like learning). Our students get projects via our platform, attend daily standups and receive (from mentors & peers) & do (to peers) code reviews. This means we have no full-time teaching staff, and that’s a huge cost-saver. Instead, we spend more effort on our underlying tech & mentors, which results in a lower variable cost per student. Important to note that even though there are no classes, our students still receive a lot of contact hours with industry experts through code reviews, help sessions, standups, etc. These hours are focused on helping our students (2-way), not teaching them (1-way) - just like you would have it in any workplace.
2. We don’t have plans to resell ISAs. We see ways to be a successful school without taking this path.
3. We believe in being focused, and Data Science[1] will continue to be our main focus. We do not plan to start any courses in other fields.
4. Our mentors (Senior Team Leads) team is hired by Turing College and is paid by the hour. We do have agreements with each of them about their weekly involvement and each of them is managing their time themselves. Because of our education model, STLs don’t have to always be available at specific hours. They mark themselves available at specific hours and we have ways to assure consistent availability coverage for help & code reviews. Because of this, our students can get code reviews any time of the week, including weekends. As for students who require more attention, that’s completely fine, and they receive help from our staff, mentors, and other peers.
[1] https://blog.turingcollege.com/data-science-job-roles-explai...
i guess it question is is there enough such students who can overcome these obstacles who are also willing to fork over X tuition? they may not need it per-se, but I think you may be discounting the value of the pre-existing networking leverage these schools have over individuals that may have no network in data science related work
The code you have on one of the quiz questions is:
----------------
a = 1 b = 2 c = 3
def calculator(a,b,c): b = a - c a = 2 return a+b
x = calculator(c,b,a) print(x)
----------------
This returns 4. But the only options you have are:
A. no output B. 0 C. 3
edit: Also, can you elaborate on this: "We're cheaper than Lambda school because we have optimized-focused learning platform & education processes. With that we can have 70% less staff to deliver the same and even higher education quality"?
Just want to understand your value propositon better as I am seriously considering to attend a bootcamp like yours.
[0]: https://strive.school
Also, our curriculum is co-created with tech companies we partner with as our Hiring Partners (we have 17 partners now). So, we know what those companies are looking for in new hires, and we adapt our curriculum accordingly; we also have their commitment to hiring our grads with our job placement program. Strive school doesn't do that.
If I understand correctly, Strive School curriculum is mostly presented in video recordings. With us, you'll be a part of our tight-knit community of peers and industry professionals, interacting daily via online calls or discord chats and working on real-world projects that our Hiring Partners are now solving.
As for the "70% less staff to deliver the same and even higher education quality ", this is enabled by letting students assess each other's work with the supervision of Senior Data Scientists. It means that we don't need senior staff for every project assessment but only for crucial ones. So by having the platform that organizes assessments in that way, we ensure the quality and the need of less senior staff. We can track how students are assessing each other, and they are doing that objectively. With our learning platform, which follows that we can react to any cheating situation instantly.
The website states that 70% of the instructors are industry experts. Are any of them known to the Data Science community outside of being backed by Ycombinator?
A good example is Dovydas Čeilutka, our Lead of Data Science, next to being ML team lead at Vinted (2nd hand clothing marketplace, valued at $4.5b), he is the President of the Artificial Intelligence Association of Lithuania and a founder of Tribe of AI, artificial intelligence learning community in Lithuania.
While Dovydas is well respected in the Baltics, we are in the progress of bringing more Data Science industry experts from global markets.
Another way we are making sure our quality is great, is by working closely with companies (Hiring Partners) by co-creating and integrating their projects into our curriculum. This assures that whatever projects our learners are working on are relevant to the market & companies they might eventually work at.
However, I’d like to know if a similar way but being a part-time/flexible student is available?
Pedagogy isn't something that tech companies are particularly known for, nor hire many experts of.
Reasons we have this are: 1. For students it's a great way to see what tech stack is the company using and what problems they are solving.
2. For companies, each coming-in student has a "technical interview" already completed and a lot more context about the company.
[1] https://blog.turingcollege.com/senior-team-leads-at-turing-c...
Knowledge is important, but an actual degree is still a major consideration, since it does still determine pay bands and such at major companies.
Is one of these wrong, or is it just a case of more than one meaning for an abbreviation?
[1] https://blog.turingcollege.com/turing-college-and-the-man-be...
Seems a bit weird to take inspiration from someone, use their name, but then malign them like that and refuse to engage with his estate.
Will you be donating any of your profits to the Turing Trust? https://turingtrust.co.uk
It's a charity set up by Turing's family to honour his legacy and to provide computers to schools in sub-Saharan Africa.
https://developers.google.com/machine-learning/crash-course/
I'm tempted to try this to add SOME data analysis skills to my existing skills.
Can we have a normal profession once again?
Are the frauds all going to start calling themselves data scientists now?
I would love that.
I just went through 5 days of whiteboard interviews for a single company.
Data science requires a very strong mathematical background. Thee are libraries and software that do take care of some of the most complicated processes, but I don't believe someone can become a good data science engineer by always relying on such libraries/software.
Hoe rigorous is the treatment of mathematical topics in the AI course you offer?
Do you teach the concepts of probability/statistics, linear algebra and calculus required for the course, together with some testing or examination relevant to the subject material being taught? Or is your approach similar to Andrew Ng's Coursera course where he does give some introduction about the maths involved without going into details because they are not required, resulting in acquisition of, at times, half baked knowledge about core concepts.