Founder of Insight here (YC W11). Since 2012 we've been running free fellowships to help PhDs transition to roles in data science [1] and more recently health data [2]. Similarly, since 2014, we have been helping professional software engineers learn and move into data engineering roles [3]. Over 750 Insight alums now work as data scientists & engineers at 200+ companies.
This past year, we've seen more highly specialized applied AI / deep learning roles emerge in the industry. We're also increasingly receiving applications from scientists and engineers who have some machine learning experience and are learning to build out sophisticated deep learning models during their time at Insight. The new Insight AI [4] program will focus on allowing Fellows with these backgrounds implement the latest ML techniques from research or contribute to open source projects under the guidance of industry leaders, then join AI teams in Silicon Valley and New York after the program. Insight AI will accept both software engineers and quantitative scientists (no PhD required).
I actually pretty interested in joining this program since I am near one of the locations! Do you guys accept other type of engineer (in my particular case Mechanical Engineering) with a background in programming?
Engineers often have a lot of experience with control theory and building maintainable systems, and individuals who have a strong math and coding background would be a great fit. Side projects in deep learning and reinforcement learning would also be a good thing to highlight when applying.
For the AI program, what kind of ML experience are you looking for?
I'm a physics PhD (working in ultrafast atomic molecular and optical physics theory), finishing up in the next few months, and I would love to transition into AI as a field. Unfortunately, I don't have an extensive portfolio of ML projects. Is this something I should beef up before I apply?
I understand you are likely much more comfortable with taking people interested in SV and NY, but for those in other areas (Denver, for example) and unwilling to relocate, is this going to kill our chances of being accepted?
For the data science (DS) program we’re very open to strong quantitative fundamentals without ML experience, as many of the DS roles are focus on analysis. For the AI program, since the focus will be entirely on getting you up to cutting edge of what’s happening in ML, it will be important to already have had experience with the fundamentals of ML. Luckily, for someone with your background, it’s completely possible to build some impressive side projects in a couple of months than can demonstrate your ability to learn ML at a rapid pace. We’re far more interested in how quickly you can learn than your current level of knowledge, so doing a side project and mentioning your starting point prior to the projects will give us strong signal that if we accept you into the fellowship you’ll be able to pick up the necessary techniques.
As an aside: it’s astounding how quickly Fellows with strong quantitative backgrounds can learn when surrounded by other really great people. We’ve had numerous mathematicians and theoretical physicists at Insight who barely touched any data in their PhDs, go on to build sophisticated machine learning data products at Insight and get hired at top tier data science and machine learning teams.
Regarding the location: the initial programs will focus on roles in the SF Bay Area and New York, so we’d like to attract Fellows who are interested in living in either of those respective cities. That said, our network has grown well beyond those cities so I encourage you to apply regardless of geographic preference and just let us know in the application where you hope to end up after the program.
Do you require that fellows to accept a job with one of the Mentor companies after the program?
The program looks really interesting and I'd definitely like to increase my skills in this area, but it's also unlikely that I would accept a job at a new company. (I would be happy to interview though - maybe a company could convince me.)
Also, I would suggest putting the important information in something besides a "white paper," like a FAQ.
Very cool! I've worked with the Insight Health Data fellows before, enjoyed the experience, and they got a lot done.
I think one thing that's tricky about artificial intelligence as a field is that it involves so many diverse pre-requisites. For example, although Caffe lets you configure a DNN with just a simple text file, when something goes wrong, the stack traces are all in C++. The abstractions tend to leak.
I know that both software engineers and quantitative scientists are encouraged to apply. On the quantitative scientist side, what level of programming ability do you think somebody needs to succeed in the program?
Likewise, on the software engineer side, what level of mathematical background would you expect someone to have coming in?
The quantitative researchers, in addition to having advanced machine learning experience, should have a strong coding background - experience with large code bases and best practices in software development. Many physicists, computational biologists, neuroscientists, etc come from this background, having worked in collaborations and implementing machine learning methods on messy real-world data collected from large experimental setups. We’ve had Fellows like this, who also in their spare time built and trained networks in TensorFlow and then built their own customized layer in C++ behind the scenes.
The software engineers coming into the program would have machine learning experience, but have not necessarily in a full-time role yet. Just like the software engineers that come into our data engineering program, it’s a chicken or the egg problem: employers want to see experience in the role before hiring for it, but how can you get experience if no one wants to take a chance on hiring you. Insight takes that chance, you work on cutting edge ML problems here, then the company has evidence (obviously combined with your previous years of work experience) to than make a bet to bring you on as an ML engineer.
Overall AI practitioners in these roles usually fall along a spectrum, having their main strength be either software engineering or quantitative research. Often companies will have experts on both ends work together to implement current research and then put those models into production.
I lead the team running the new Insight AI fellows program in Silicon Valley. For the past year, I've been providing guidance and mentorship for Insight Fellows doing deep learning projects as part of our data science program. My background is in applying current deep learning research to large scale audio and video data sets to understand animal behavior and cognition. I'm happy to answer any questions.
Hi, for PhD students, when is the recommended time for applying to this program? For the data science program, I believe it is 1-3 months before degree completion; is it the same case for the AI program? More specifically, I'm on track to graduate in 2018, but interested in looking for internships/bootcamps during summer 2017.
You need to be able to start your new full-time AI role within about 2-3 months of completing Insight. So we recommend applying within the last few months of your PhD, which will allow for a smooth transition from your program into your new role.
This is a very interesting program and I imagine there will be lots of applicants. So this leads to my question, how many fellows will be accepted to each session? Also, for those of us who see this as a life-changing opportunity, what would you suggest we do to improve our chances of getting accepted and successfully completing the program?
Hi, would you consider graduates from non-CS fields? I will very soon have a PhD in Civil Engineering where I do statistics and Machine learning for solving weather-related problems. Have good experience with R and Python.
Yes, absolutely, one of the goals of the program is to help researchers with strong quantitative backgrounds and experience in machine learning enter the industry, despite not having a background traditionally associated with ML.
I've been following Insight since Jake was subleasing office space from us for his first class of fellows (there were six of them occupying ~500 sqft. Their Palo Alto office alone is >14,000 sqft now, I think), and from the very first class it felt like a very special program.
I've come back to both mentor and hang out with their fellows every single batch since (I think I've been to 15 now), and it continues to amaze me just how incredible the people are, and how cool the community is.
If you're thinking of a career in AI and you have some of the fundamentals they're looking for at https://www.insightdata.ai, you'd be crazy not to apply.
This is exciting for me personally since I've been working on building a portfolio to transition into a position where I can work on cutting edge DL/RL work, so I will certainly apply, though July feels so far away .
Can you say which companies you have lined up to hire people in NYC?
P.S. If anyone is looking to hire a research engineer before August of next year, my email is in my profile... :)
Since the New York session is a bit further out, we're still finalizing the mentor companies. That said, for our NYC data science and engineering programs, companies hiring Insight Fellows include Facebook, Bloomberg, Capital One, New York Times, Memorial Sloan Kettering Cancer Center, and dozens of others. We expect many of these companies plus other NY AI teams to be actively participating as we have already received significant interest.
The litmus test for me on whether we're adding adding value as an education company has always been: are there Insight Fellows who get rejected from companies X,Y,Z prior to Insight then get offers from X,Y,Z after Insight? From the very first session through to today, we have numerous examples each session of this happening.
A recent example was a Data Science Fellow who was a physics postdoc at Lawrence Berkeley National Lab prior to Insight. Right after his postdoc ended, he applied to half a dozen bay area tech companies (all the usual brand name suspects), got rejected from all of them. He came to Insight and during his fellowship built a video scene segmentation & object detection project with a YC startup. After Insight he got an offer from every one of the companies he previously got rejected from. He went on to accept an offer on the LinkedIn data science security team (which is led by another Insight alum).
We’ve seen this happen time and again on the software engineering side as well with our data engineering program. A Data Engineering Fellow prior to Insight has a generalist software engineer experience but a passion for big data, wants to do big data full-time, but no one will take a chance on her/him. At Insight they build a sophisticated data pipeline on AWS, while being mentored by leading data engineers, and then the same companies previously rejecting that Fellow for data engineering roles make offers because they now have the evidence they need that she/he can solve the types of specialized problems the company is facing.
In your experience, how much of the interviews for an AI / DL role consists of classic CS algorithm puzzlers, compared to a regular software engineering interview at a place like Google / Facebook?
I think you're partially right about these programs picking 'winners' and ushering them into jobs they could have done without this program since the actual work here boils down to a 3 week project.
However, it can be pretty hard to even get an interview if you don't fit what recruiters pattern matching against, so in some sense this is helping companies make smarter recruiting decisions.
I must admit that my interest in this field has peaked in 2015 (say a mid 2014 - mid 2016 cycle) and that I'm now only keen on applications for my industry. It's a good sign, though: DS/ML/AI alphabetisation is progressing very well and a lot of people and outsiders like me are starting to use the tools for their own agenda.
This past year, we've seen more highly specialized applied AI / deep learning roles emerge in the industry. We're also increasingly receiving applications from scientists and engineers who have some machine learning experience and are learning to build out sophisticated deep learning models during their time at Insight. The new Insight AI [4] program will focus on allowing Fellows with these backgrounds implement the latest ML techniques from research or contribute to open source projects under the guidance of industry leaders, then join AI teams in Silicon Valley and New York after the program. Insight AI will accept both software engineers and quantitative scientists (no PhD required).
[1] Data Science: http://insightdatascience.com
[2] Health Data: http://insighthealthdata.com
[3] Data Engineering: http://insightdataengineering.com
[4] Artificial Intelligence: http://insightdata.ai