Last Friday (16th August 2019) we had AI Singapore’s (AISG) AI Apprenticeship Programme (AIAP)™ batch #2 graduate after 9 months of training, which consisted of 2-months of intense self-directed learning and ten 3-weeks sprints which they delivered a minimum viable AI model to our 100E* project sponsors. With this batch, AI Singapore Industry Innovation team have solved our Singaporean AI talent crunch issue and have managed to hire a strong core team of Singaporean AI engineers to accelerate our AISG mission.
This article will share the AIAP™ journey and how we built a strong Singaporean core of AI engineering talents.
Figure 1. AI Apprenticeship Programme batch #2 graduates. Photos includes our special invited guests that evening: Mr Aurelien Geron and Mr Jared Lander.
The AI Talent Crunch
Artificial Intelligence(AI), Machine Learning (ML) and Deep Learning (DL) are the hype these days, and nearly every company and start-up is either planning or has embarked on an AI project, and/or changing their business plans to include AI and ML.
Since we started the AISG programme back in June of 2017, the number one comment we get from companies here is “I cannot find Singaporean AI Engineers!”
This is a valid comment as we experienced it ourselves when AISG wanted to hire AI Engineers to work on the 100E AI/ML projects. We received nearly a hundred applications when we advertised for the AI Engineer position back in August 2017, and only a handful of Singaporeans applied, however most did not have the right experience or required skills. This was not a sustainable position for AISG and we felt it would severely impact our 100E deliverables, and as the Singapore’s national AI programme office, we need to have a strong Singaporean core of AI Engineers.
The AI Apprenticeship Programme
With an initial AI engineering team that included both Singaporeans and experienced SPR and foreigners with expertise in AI and ML engineering and decades of grooming young engineers, we developed the AI Apprenticeship Programme, a 9-months hands-on apprenticeship programme where candidates get to spend time to deepen their AI/ML skills in 2-months with a carefully designed curriculum which provided the necessary knowledge to take on an AI/ML engineering project followed by ten 3-weeks sprint to deliver a minimum viable AI model.
Figure 2: The AI Apprenticeship Programme
During the initial 2-months, we do not teach in a formal classroom settings (those days are over), but instead, our AI mentors guide the apprentices to learn and do the following:
- Agile methodology
- Software engineering
- Containers and dockers and their deployment
- Big data with Hadoop/Spark/HPC
- Supervised ML
- Unsupervised ML
- Deep Learning
Basically, each week, the AI mentor will say: “This week we are focusing on unsupervised learning, here are the materials and resources, and please show us your project on Friday!”. The apprentices are expected to use what they have learnt to develop an AI model in an agile manner with dockers and deploy them including all the ingestion pipelines and data processing required.
We are able to do this because the candidates accepted into the apprenticeship programme have proven themselves to be self-directed learners. This is one of the traits we look for when we bring them into the programme, and it allows us to run at a quick pace. There is no spoon feeding here, no classrooms where you attend tutorials or lectures by the AI mentors. Instead, often, it is the AI Apprentices themselves who self-organize and conduct their own workshops to share the latest AI papers and teach each other.
From week 6 onwards, the AI mentors will have a sense of some of the strengths and particular challenges of the apprentices, and it is also the time we start to share and discuss the 100E projects that needs to be executed in the next 7-months. The 100E projects are listed on a board, and the apprentices can indicate the projects they are interested in. We try our best to assign the 100E projects based on their selection, but often we need to allocate the projects based on the apprentices strengths, background and future aspirations.
The AI apprentice in the course of the 9-months programme would have built a Minimum Viable Model (MVM) in ten 3-weeks sprints, and deployed a Minimum Viable Product (MVP) into production in collaboration with the project sponsor’s engineering team. Our 100E outcome is a deployed AI model, and to date we have successfully deployed all of our 9-months 100E projects AI models.
One of our best practices in contrast to a typical internship model, is that we require all apprentices to be based in the AISG premise for the whole 9-months including the period they are working on the 100E project with project sponsors. This allows them to easily access our AI mentors and other apprentices within the cohort. This leads to accelerated and shared learning between the apprentices, but more importantly, they build very strong bonds and friendships that will last them beyond the 9-months programme.
Getting into the AIAP™
Figure 3: The AIAP™ application process
The AIAP™ accepts only SINGAPOREANS who pass both our technical assessments and interviews. It used to be that we require candidates to have a university degree. That have been relaxed to a minimum diploma requirement starting from AIAP™ batch #4 onwards starting in September 2019. AISG currently takes in 3 batches per year, and we typically get 120 – 160 applicants per batch, and only around 10%-15% gets admitted into the programme.
In the AIAP™, we are looking for candidates with the right skills, expertise and attitude. They typically come with intermediate to advanced programming skills in Python, and have done self-directed learning from books, and various online AI/ML MOOCs. What they lack is experience working on a real-world AI problem. This is what the AIAP™ provides.
Who are our Apprentices?
In selecting the apprentices, we do not consider which discipline they studied in or what their academic scores were when they were in school. There is no correlation between what they studied in school and how good they will be as an AI/ML engineer or scientist. Some of our best AI engineers studied economics, business administration, industrial/civil engineering and biology.
From the 4 batches we have admitted, we typically have the following in terms of academic backgrounds:
Our apprentices are not students. They are either fresh graduates or working Professionals , Managers, Executives and Technicians (PMETs). The 4 in-takes we have so far have the following typical proportion:
There is also a consistent ratio of around 12%-15% female apprentices to males per AIAP™ batch.
In less than 15 months since we started the AIAP™, AISG have solved our problem of “lack of Singaporean AI Engineers”. We retained 2 out of 13 apprentices from batch #1, and 9 out of 26 in batch #2, and released 28 AI Engineers into the industry. We have another 18 undergoing training in batch #3, and 18 new apprentices in batch #4 joining us in mid September 2019. The AIAP™ will continue until 2024, and we hope to train 400-500 AI engineers by then.
We have found a sustainable model to build up our Singaporean core of AI engineers for ourselves and importantly also for the industry. We are not a big team in absolute numbers of AI engineers, but nowhere in Singapore will you find 40-50 Singaporean AI engineers in a single room solving real-world AI problems for companies!
So when you visit AISG, and step into our engineering hub and ask one of our engineers “Excuse me, are you a Singaporean AI Engineer?”, the answer will likely be a YES!
* 100E – To accelerate and help Singapore based companies and startups in this AI revolution, AI Singapore (AISG) developed the 100 Experiment (100E) programme. Under the 100E program, companies (Project Sponsors) comes to us with their AI problem statement, and if suitable, AI Singapore will co-create the solution on a 50-50 funding model. AI Singapore will invest up to $250,000 and companies will invest similarly and all the cash and resources will be directed to an AI engineering team assembled by AI Singapore with the local Universities and research institutions. Note that no cash funding is provided to the company.