PyData London Meetup, a meetup for which it is notoriously difficult to obtain a ticket for. With a significant waiting list each month PyData London is a meetup for which you have to be very quick to confirm your attendance. Having previously met the organisers and attendees at the pub after a previous event (as is Meetup tradition) I knew that the quality of the community and talks would be of high quality and I was in the right place to continue my Data Science education. If you fail to get a ticket to the event but still want to meet the community follow the PyData London Twitter feed to know the location of the pub they attend after the meetup.

The Location

The meetup doesn’t move. Why would you when there’s free food and beer? This is because the meetup is continually hosted by the charitable team at AHL. This allows the organisers to spend more of their energy on other aspects of the meetup rather than looking for new locations each month. Because of this the meetup grows in quality each month.

But, because the venue doesn’t change the meetup now regularly reaches maximum capacity within a couple of days of being announced (!). Unfortunately this means that only 230 people can attend each time and people do miss out. The organisational team know this and almost all talks end up online and shared to the meetup group. A significant group also go for a drink at a nearby pub after. You miss out on only a little by not attending and it’s to the organisers and speakers credit that that is so. Although because of this system the meetup will live longer, the hard to attend aspect making sure that the act of getting a ticket will be seen as a more significant event and the ratio of actual attendees to places offered should remain high for the foreseeable future.

The Main Talks

Ju Lui — Predicting Titanic survivors with machine learning

Live coding for 45 minutes in Python, his second choice language? No problem for Ju Lui. From a beginners perspective this was a perfect introduction to Machine Learning. With the first half being devoted to visualising and understanding the data of the Titanic survivors through use of the pandas package everyone in the room was able to remain on the same level as the speaker. All live coded, no errors. Strong.

Having got so far in my statistics education to reach multiple variable linear regression analysis the final third of the talk creeped beyond my understanding but not by so much that I disconnected from the talk. Having understood the data before reaching this step I could understand 90% of what was discussed and understand the supervised learning taking place. Overall, the talk was high-quality, educational and entertaining. Seeing as the talk was 95% live coded the bar has been set high for subsequent talks. Relevant GitHub.

Egor Kraev — Asynchronous and streaming Python for real-time analytics

Egor Kraev, having spent 10 years as a quant at two major banks, is now Head of AI and Founder at Dagon Analytics. The talk focused on another very significant part of AI. The data systems allowing the data to be structured and accessible so the algorithms can complete their tasks easily.

The talk focused on discussing a push approach within a data framework and then changing it to a pull approach powered by asyncio and Kafka. The logic behind the decision to change from a push to a pull system was well explained and visualised. The talk concluded with a quick demo showing the system running on a laptop. The talk was a higher level than the one before but still allowed for those without the relevant experience to understand and learn. Great talk again!

Lightning Talks

5 Minutes max. Go!

Matti Lyra — Hands on with GloVes

Matti Lyra came over from Germany for both this talk and to inform the PyData London community of PyData Berlin, another conference after PyData London. To communicate a useful amount of information to a crowd in under 5 minutes is hard. Matti did it in 2. It went straight over my inexperienced head so I’ll leave you with a description and a link to the appropriate GitHub utility.

Distributed word representations in the form of word2vec and GloVe have become very popular in the last 5 years. While the Stanford NLP group has published their trained GloVe vectors, the data format isn’t exactly convenient. I’ll present a small utility I wrote `GloVe2H5` that makes accessing the pre-trained vectors much easier.

Adam Hill — Building a graph model of corporate ownership data to uncover potential corruption

Adam Hill stole the spare time from Matti’s talk to allow him to discuss how DataKind are using the publicly data from Companies House to understand corporate ownership. Originally a 15 minute talk, the act of condensing the information down to ~7 minutes didn’t reduce the significance of the work. By being able to visualise and analyse ownership graphs of companies a lot of insights were able to be studied regarding tax havens and partial ownerships. The results were communicated to the Government which publicly acknowledged the good work and are now using it in investigations. Slideshare available here.

Celine Boudier — Dashboard for Code For Life education tools data with Google Data Studio Beta

Celine Boudier introduced us to a non-profit initiative from Ocado called ‘Code for Life’ with the aim to create free, open-source games that help children learn computing. The first game Rapid Router is already in action! Celine wanted to acquaint us with Google Data Studio. Celine and the team use the software to visualise the data that they collect and make it more communicable to other people. These easy to produce visualisations allow them to make sense of the behaviour on their site and because it is web based it helps them solve problems that choosing to use a Linux based system can bring (no Tableau). SlideShare.

Round Up

The meetup was great and well organised with both the topics and the community matching each other with their high quality. I met many great people during the meetup and I look forward to meeting more at further events!

VERDICT: Sign up to the community and go! Beginner or Experienced.

Next Event: PyData London Conference. Buy Tickets Here. No meetup in May.

As of 7/4/17 I am not yet a Data Scientist nor an APM. My grand quest is to be a Product Manager with good Data Science knowledge and/or background. If you would like to help me, even just a little with that quest please connect with me on LinkedIn or contact me on Twitter @robwin. I’ll answer for sure and it’ll be massively appreciated. I’m looking for a team to learn from and produce value with.

PyData London Meetup April 2017 Review, by a beginner. was originally published in TDS Archive on Medium, where people are continuing the conversation by highlighting and responding to this story.

This is how I’m becoming a Data Scientist without experience in the traditional sense. The market for Data Scientists is screaming out for 2+ years of experience but is correctly sceptical of persons without it. To solve this problem I need to take a different route to the role. I’ve been talking with a number of Data Scientists and Tech Recruiters and taken onboard all that they’ve said. Through the huge amount of information that I’ve been given I’ve been able to come up with a plan which I am in the middle of testing. All of the evidence says that my plan is working.

In Data Science terms, I’ve produced a model from a large amount of data and the data on which I am now testing it currently fits the model.

The base that I’ve built

For this plan to apply to anyone else I need to explain the position I was in before I started to try to become a Data Scientist. Also, should you not want to become a Data Scientist the overall plan should be adaptable to any recruitment market where the market isn’t in your favour.

I’ve interned at multiple Tech startups. The experience from these internships allowed me to learn about myself and learn in what role I would be best utilised. They also enabled me to build an initial network which has made my quest a lot easier. I also have an MSci in Physics; a quantitative degree gives me a great standing and when job hunting an appropriate degree will always help. Those of you without a degree, some of the most talented people I know don’t have a degree. They work hard to make themselves stand out and they’re doing well, these things are possible.

That covers experience and hard skills. But soft skills — people skills, team skills and other types of skills are a different beast. I believe that this plan should work irrespective of soft skills but good people skills will speed up the process and make you stand out. If these skills are weak they should be worked on. How? Reading novels is a great start, Stoner is a personal recommendation.

The Plan

Step 1 — Network, Listen and Learn

How can you know you want to do something without experience or knowledge? You’ll be asked that question first. Experience can come through networking and practice (Step 3) but knowledge is the first priority. Knowledge can be found online (Step 2) but useful knowledge will come from people.

There are 2 types types of people you want to find and learn from. Data Scientists and Recruiters. Ask people in your network to introduce you to them. Ask to talk to them and learn from them. Ask how they became Data Scientists and how you can become one too. Ask recruiters how to stand out, not just in a pool of Data Science CV’s but in the whole recruitment market. Listen and Learn, that’s the most important part. Ask them to introduce you to further people. If you’re truly interested and you can put that across in your conversations they will want do that.

If you can’t find these people through your network there are other ways. This is the reason you’re on LinkedIn. Search for Data Scientists near where you are based and send them a message. Some will ignore you, many won’t. The key is to do as much as you can of the heavy lifting yourself.

Meetup.com was designed for people such as yourself. In London there are a number of Data Science meet-ups (PyData, Data Science Festival, and many more) which I attend looking to learn. At these events I have met some great people and I learnt a lot.

Then there’s the secret sauce. Apps and services that connect individuals. Take Tinder, remove the hedonism, make it professional. You’ll end up with Shapr. But Shapr is more noise than signal (much like Tinder). The real secret sauce is JamieApp (only currently for Tech and Finance). That’s how I’ve met movers and shakers.

Step 2 — Learn

Fill in the gaps of your knowledge. Begin to develop the skills needed for the job. For Data Science this means coding, statistics and communication. You can learn the hard skills online easily and cheaply, no need for another stint at university. Udemy (when on sale) is unbelievable value for money. Udacity’s Nanodegrees are great. DataCamp is an excellent place to start. Alas, there are too many options and that makes it difficult to start, I was only able to come to a decision through someone else’s hard work.

Learn how Data Scientists communicate. How they articulate what they learn to people who know the inside knowledge and those who don’t. You can do this by signing up for appropriate newsletters (Data Science Weekly) and searching for appropriate articles on places like Medium, Google, Reddit and LinkedIn, KDNuggets. Learn to visualise well. Excel graphs don’t cut it anymore. Tableau and Carto have public versions enabling people to create great visualisations. This talk is also useful.

Listen to Podcasts. This is how Data Scientists continue to learn and solidify their knowledge by hearing different viewpoints. Some great ones are Data Skeptic, More or Less and Partially Derivative.

Step 3 — Practice

You can learn to code. You can learn the Data Science process. You can have a PhD in Statistics. But without experience you’ll struggle on the job and need guidance, you may not have that. How to get experience outside of a job? Kaggle, recently acquired by Google, is a place where Data Scientists congregate, practice and learn from each other. With datasets and their associated challenges you’ll learn to be a data scientist beyond what you can learn from the online courses in Step 2. A similar place is HackerRank. Anything that you produce post on GitHub and put your GitHub on your CV. Learn the Git process, how it works and how it will apply to you. Don’t worry if your code is bad, the more you practice the better you will become.

Step 4 — Apply

I have written previously about why Junior Data Science jobs don’t yet exist. In fact, they’re starting to appear out of the woodwork. But that doesn’t mean they’ll be as high quality as a structured Grad Scheme. The world’s lack of experience with Data Scientists will mean that, until they are given a good direction, their skills can be and will be easily wasted.

But when that right role comes along you’ll want to be ready. And being ready comes through practice. You can practice the role in Step 3. But if you’re not ready for interviews Step 3 will have been a waste of time. Talk to the Data Science recruiter that you would have found in Step 1. Ask them if you can practice with them. If they agree, they’ll help you be ready and you’ll help them too by helping them practice and learn the right questions to ask.

Some Motivation

This isn’t the only way. This is a plan that I’ve put together from the advice I’ve received. It’s a difficult, time consuming plan that could probably be optimised. In a mature market a graduate would be likely earn their stripes cleaning data. But without experienced Data Scientists and Data Engineers creating useable, well structured data there is no need for the typical graduate scheme placing graduates in pure Data Science roles, yet.

Besides, if you do this, you’re a pioneer. School and University were guided, there was a system in place designed by others to get you from A to B. The path was well trodden and optimised. Going from University to Data Science is a path that is yet to be trodden. I have learnt, and am continuing to learn, a lot about myself during this journey and I hope that if you were to go on a similar journey the same would happen to you. Through this plan I think that I’ll be ready and raring to go when I get to the first rung of this career ladder.

As of 29/3/17 I am not yet a Data Scientist. If you would like to help me with that quest please connect with me on LinkedIn or contact me on Twitter @robwin. I’ll answer for sure. I’m looking for a team to learn from and produce value with.

It’s taken me 18 months since graduating to decide what I want to do first. I’ve tried a couple of things; designing solar farms, selling experiences in the mountains, taking people scuba diving and being a generalist at multiple Tech start-ups. The journey has been long but discovering Data Science has very much excited me. This mountain of data that we are producing can be and is being used. And it is being used well.

In the noise produced from all of this data there is a signal. Stitch Fix are finding it and by using it they’re making sure you’re wearing the right clothes and feeling confident. Transport for London are learning how to optimise your commute at the moment. Imagine that, a better commute?

Mind, this has been happening for a while. Your loyalty card from Nectar or your John Lewis Credit Card has allowed you to receive coupons in the post every couple of months. These aren’t random, there’s some serious science in play. Each coupon is connected to your spending habits, making sure they’re useful for you. To achieve this they’ve employed Data Science and they’ve employed it to great success.

So, How do I become a Data Scientist?

As I’m learning at the moment, it’s difficult. As more and more data is being produced and stored by companies all around us we’re only just starting to realise the value of this resource. And much like oil, gas and other resources data is another resource for which we’re finally building an engine to use it.

Companies know this and they’re hiring, a lot (according to Hired). Data Scientists are wanted and needed everywhere. But there’s a catch…

Enter the CEO’s internal dialogue.

“Ok, I want a Data Scientist.”

“Why?”

“We’ve got all of this data that we’re storing. Apparently there’s value in it, and I believe it.”

“Fine. Let’s say there is. Who’s going to play the prospector?”

“That’s what a Data Scientist is for?”

“What are they going to do?”

“We’ll hand them our dataset. They’ll optimise the business and play oracle. Through them we’ll always know what to do.”

“Sounds great! Let’s get a Data Scientist then”

What’s the catch? Experience. The Data Scientist that they want is someone who will know what to do. They can simply give them data and the capital to build a team and in 6 months to 2 years they’ll produce a profit.

This shows the first problem with the overall market for Data Scientists at the moment. Data Scientists are being employed without direction. A CEO/CTO will build a data science team without an idea of what they can do and just expect results. This particular C-level lacks the experience to get the most value out of a Data Scientist. The solution? There needs to be an intermediary, someone who will give direction, the correct direction too. That someone will have experience.

Back into the mind of the employer.

“Ok. So I want a Data Scientist with experience. I also need to have an idea of what I want the Data Scientist to do.”

Calls the CTO and builds a strategy for the data. (Easier said than done, probably involving more people too…)

Once complete, they call in the Head of Talent Acquisition.

To the Head of TA, “We need a Data Science team. We need about 5 people with 5+ years experience.”

Head of TA: “OK”, whilst thinking internally “Doesn’t ask much…”

The second problem now rears it’s head. Supply and demand. Demand outstrips supply, a lot. Experienced data scientists are rare and when found they aren’t cheap.

“So how does this apply to me, the graduate?” you ask the gods of capitalism. They reply a simple “Good Luck.”.

A healthy graduate market with a stable recruitment funnel (what you’ll experience with finance and consulting) cannot exist without other, more experienced people offering guidance and training.

So, if you, like me want to be a Data Scientist there’s an uphill struggle ahead. There’s no obvious path (without increasing your debt). Companies such as Accenture and Deloitte pay a lot of money to make their hill is more visible than others so you know that they’re an option when you leave university.

The majority of companies don’t yet need graduate data scientists. They need experienced data scientists. The only graduates I’ve seen working in such a capacity have been working for Channel 4 (and they’re doing some cool stuff).

Some Motivation

But this doesn’t mean that you should give up on the idea. In the graduate market as a whole the supply outstrips demand, especially for great jobs. If you want to stand out in both the graduate employment market and the general employment market you need to shine and figuring out a way to become one of the first graduate data scientists is a way of doing that. You’ll honestly be able to call yourself a pioneer.

Mind, if you’ve got an applicable PhD you’ll find it easier. Pivigo’s S2DS is an obvious starting point.

If you want to know how I’m planning on becoming a Data Scientist watch this space…

As of 27/3/17 I am not yet a Data Scientist. If you would like to help me with that quest please connect with me on LinkedIn or contact me on Twitter @robwin. I’ll answer for sure. I’m looking for a team to learn from and produce value with.

What’s your favourite digital product and why?

Definitely Intercom, a mashup of customer development, analytics and more all in one package. It’s a great product that has allowed me to produce some amazing solutions to business problems with so little hassle that I’d recommend it to anyone wanting to interact with their customers.

Intercom was built with to enable easy communication with customers and they’ve done it well. Interacting with and learning from a customer is a seamless experience whether they’re using your product there and then or not. Through an unobtrusive speech bubble on your page you can be connected with your customers at any time.

This experience needs to be seamless. It’s a clever marketing channel in itself. When customers have a great experience using Intercom to communicate either on the website or by email there will likely be some explorable Intercom branding somewhere.

Now the analytics side of Intercom. This isn’t talked about much but it’s the best part. It’s understandable that people don’t talk about it though, people don’t like knowing that they’re being tracked (but we accept and ignore it, especially if we can’t see it in action). The data it collects and the tools they’ve built for that data are easy to use and can enable people to do great things with the system. We were easily able to analyse the way our product was used. If our customer left and never returned or whether they finished onboarding was able to be tracked and understood easily.

With the ability to automatically sort users into segments depending on their usage you’re given a lot of power to do some amazing stuff. We built a mini CRM to guide communications with our clients which paid dividends instantly. I wonder what other teams will have built?

Oh and our developers loved it too. Easy and quick to implement so it was a great use of engineering time.

What’s one thing you would Like to improve about our Product?

Difficult, Skyscanner is a good, easy to use product evidently built by some great people. But enough flattery.

I’d make one improvement and it looks as if they’re already working on it. Their dream dynamic.

On the front page, close to the flight information inputs (a difficult problem to solve in itself, being asked 6 questions at once without an allowance for ambiguity is going to create problems) there’s a link to a map which would, should it be more prominent (as it is further down the page), reduce the number of less potential customers scrolling. Here you’re allowed to explore and learn where you could go. This type of product enables people to dream.

But even here I think there’s too many choices, it should be guided further or you’ll get lost in the sea of information. This guiding, I think, could be achievable through either an unobtrusive search dynamic or chatbot making sure people are looking in the right places.

Ideally this search dynamic or chatbot would be able to work with ambiguous questions and answers. If Skyscanner asks an open question such as “What do you want to do?” I would hope to be able to reply “Ice Diving”. Or if I’m asked “Where do you want to go?” I would be able to reply with “East” or the sun emoji 🌞.

This ability would require a lot of data and the data be used well with some difficult natural language processing. Thankfully Skyscanner is a company that loves data and I don’t think this is beyond them. The idea is hard and a lot easier said than done, but I certainly wouldn’t shy away from it because of that.

What was your favourite project at university and how would you apply your learnings to our APM role at Skyscanner?

Definitely my Physics MSci project. Designing and testing a new solar cell structure made from nanotubes was interesting, testing and a lot of fun. But how I learnt to navigate complex team dynamics would be the most valuable skill I would bring to Skyscanner.

I was working closely with Professors, Post-Docs, PhD candidates and other masters students. Their ways of working were entrenched and they were working hard with tight deadlines. Straight away it’s easy to be viewed as simply an obstacle, taking up space in the lab and distracting them a lot of the time. Thankfully the team I worked with for 3 months were accommodating and helpful and I wouldn’t have completed a first class project without their help. But had I acted differently and without respect to their work I expect the story would have been very different.

What challenges do you think online travel brands will face in 5 years time?

“Why Fly?”

The business market will be asked this question much more when Mark Zuckerberg perfects some of his early VR projects with Oculus and Facebook. The lack of a requirement to need to be physically in the room for a meeting and instead just being in a virtual room for meetings will reduce business flight sales significantly.

The only solution to this problem I can offer for now would be to focus on creating some great marketing campaigns explaining the value of true face to face and in person communication. Especially in sales.

To conclude, I enjoyed answering these questions and should I be granted an interview I’ll be more excited than nervous as there is a high possibility that I’ll be asked equally good questions. If not, Skyscanner will remain in high regard with me because of the intelligent questions they’ve asked.

One of the problems that I often encounter is my reply to the competency based questions that I am often asked. I’m simply not great at answering them.

This is likely due to a lack of interview practice. But I’m looking for immediate results and that will have to involve skipping the practice part. These are dream roles I’m applying for and I’ve got to do my best to guide the interview before it even starts. One way of guiding the interview is making sure to be to be asked questions that I can answer well. To do this I’ve got to think outside the box and ‘hack’ the system.

There seem to be multiple ways of doing this. The most obvious is CV design. Making particular sentences more prominent than others can guide the eye of the reader when they glance over the CV. But if the CV is not even read beforehand this attempt is made redundant. Another way, which I am now unapologetically attempting, is acknowledging good questions. By giving feedback on the system, the system can then change to be more successful and improve the candidate experience, or that’s what I’m hoping.

But this case is slightly different. Sometimes you can be asked these good questions that I aim to be asked before an interview is even arranged. This makes both the company and their respective recruiter stand out from the rest and be valued by the applicant.

A good example is my Skyscanner application. I’m currently applying for the Associate Product Manager role and I’ve been asked 4 good questions which I’ve found easy to answer:

Tell us about your favourite digital product and why.

What is the one thing you would like to improve about our product?

What was your favourite project at university and how would you apply your learnings to our APM role at Skyscanner?

What challenges do you think online travel brands will face in 5 years’ time?

Why are these questions good? It may be a problem related to my inexperience, I’ve only been asked questions appropriate for grad schemes in the past, but I’ll ignore that and analyse them nonetheless.

1. Tell us about your favourite digital product and why.

This question asks me about myself and positive interactions I’ve previously had with digital products. It’s been said hundreds of times, people like talking about themselves and I am no exception. Talking about a product that I have enjoyed using allows me to think about a positive past and reminisce. Once the particular product is chosen (the hard part) the question is easy and enjoyable to answer in which I talk about positive experiences I have had.

2. What is the one thing you would like to improve about our product?

The question is about both myself and Skyscanner and allowing me to think about a hypothetical future where we work together (with the aim of removing the hypothetical part). This hypothetical future is one where I would be working in the role that I want and making changes to perfect a product without the real world limitations that I would experience when attempting to achieve my goals. Although this is a virtually unobtainable dream (that a good PM could achieve), everyone enjoys dreaming and so this question is once again enjoyable to answer.

3. What was your favourite project at university and how would you apply your learnings to our APM role at Skyscanner?

I once again return to the past and I am reminded of the positive interactions I had when at university. Having once again reminisced, this time to a previous project that I’ve enjoyed, I take what I have learnt to that hypothetical future where I’m working in the role. So I would apply these positive interactions to the hypothetical future to ideally create a positive hypothetical future. Once again, questions about positive futures involving myself are enjoyable to answer.

4. What challenges do you think online travel brands will face in 5 years’ time?

This final question allows me to think of the bigger picture, identify a problem and theoretically solve it (the physicist part of me adores this!). To think of grand situations is an exciting part of business and is one of the main reasons I swapped from academia. To then attempt to solve this huge problem that is sitting on the horizon with only words is nice and a lot easier than doing in actuality.

So, I’ve enjoyed answering these questions because they’ve made me think of past, positive interactions and ideal hypothetical situations in the future with a focus on tech. Because of this these questions have had the effect that even if my application is unsuccessful I will continue to hold Skyscanner in high regard.

Unfortunately I’m asked to supply these answers in a cover letter which means I have to reduce their scope to save valuable space. Also, adding pictures to a cover letter in the main body of text is an innovation which I will not lead. So I’ll publish the long form answers here soon…