The motivation to write this post is to provide a comprehensive reply to the questions I have been asked six times just today (and to save some time by not having to type individual replies).

>What class did you take to get started in data analytics?

>How much did you spend on the classes?

>Can a person from a non-tech background do fairly well in data?

>Will I be able to learn data and programming if I don’t have a “math” mindset?

>Was it hard to get your first job in data?

>How much were you paid at the start?

About myself: I graduated with a BA in English Philology, I worked in marketing for about 7 years, I have never programmed before. In 2016 I hit a very low point in my life, I had no money, burned through my startup, lost my job, went through a divorce, and was raising two kids on my own with no idea how to pay the rent next month. My main motivation to move to tech was to make enough money to support myself and the kids in the Bay Area. I was scared about a future without a good career. And I was even more scared of failure should I be brave enough to transition into tech. Because of this fear I postponed learning data and programming for about 2 years.

At 36 I thought I was too old to learn programming. At 40 I realized I wasn’t.

Hopefully this post can bring some clarity to those who are just making this sharp career turn and are a few years behind me. I love how Rachel Thomas put it “You are best positioned to help people one step behind you.” If you are just starting, this post is for you.

Let’s be clear about classes and courses and how those can help you. I don’t think there is just one magic class that will give you everything you need to know. If you don’t have a CS or engineering degree (I didn’t!) expect to take a variety of classes on the same topic. Why? You will find out that every teacher explains the material differently and listening to the same content from three or four different instructors can help you understand it throughly. I advise not to focus on finding the best course, but instead sign up and attend a few courses in parallel. You can do classes on Coursera, at a community college, on Udemy, or even do tutorials on TeamTreeHouse or DataCamp. During the first two years attend as many classes as you can, ideally every day. For example, I started with R specialization on Coursera. I never use R in my work now, I use Python and Tableau, however, learning R gave me some basics of data analytics and allowed me to orient myself in the field.

You don’t need a large budget for learning python and data anlaytics. Most of Coursera classes are at $49/month (at the moment of writing). Data Science in community college cost around $1,300 for an academic year. Some classes and bootcamps are free or have scholarships. At the end of this post I will list the classes and courses I took.

I think that you can move to data analytics from any background, at any age, and at any stage of your career. And you can be succesful! However, please remember that it will take time to advance in the field, it will not happen after a six-months bootcamp. I think that you can get an entry level job as data analyst after one-two years of evening/week-end studies and some Github code to prove your skills. If you can, while you study apply for a job that is not technical but bordering with tech and from where you can grow into a more technical role. Feel free to check jobs at Accelerate Change (the company where I started my tech career). I love their culture, kindness, and focus on constant learning supported IRL (allocation of learning time and budget).

Can you succeed in a data career if you don’t have a “math” mindset? I don’t think this was my case and I can’t talk outside of my own experience. I always loved math and statistics and I welcomed the idea of working in tech, I just was very apprehensive that I wouldn’t be good enough. On the other hand, I don’t think there is such a thing as a “math” or “not-math” mindset. If you can do your household bills and understand the sale percent off on an item then you understand numbers, and with a little push you can understand data. Is it possible that our fear of failure in a tech role convinces us that we don’t have a “math” mindset? I was always against sabotaging myself, thus I would put the whole idea of a “math” mindset on a back burner and give it a try.

My main advice: don’t quit. Even if you don’t understand the material and lose faith in yourself, just keep showing up to class, keep googling answers to the questions you have, keep asking questions in the class, keep doing your homework. If you don’t quit during the hard moments (first two years), I am sure you can be good at it and find a job.

The hard part about the first two years (specially for those coming from a non-technical background) is that you seem to live in the middle of a black hole. The problem is not that you don’t know something. The problem is that you don’t know what you don’t know. And this makes you feel like living in the darkness with no vision or clarity. Take my word for it, just keep showing up. With discipline it will get better and eventually you will see the light and things will make sense. If you lose hope, may the discipline be the light to guide you.

As we joke with a fellow data scientist, you are a good data director if you can google the correct answer in under 4 tries. There is a lot of truth to this joke. In many cases you will master a narrow skill set that your job requiers. You don’t have to memorize the rest (it is actually impossible), what you need is to understand the problem, know what tools are available, and be able to search for what you need. Ability to formulate the question correctly is THE skill to master. Aim at that!

Let’s get to the job part now. Was it difficult to get my first job in data? No, it wasn’t, at least in my case. However there is a caveat. I started working at Accelerate Change as a content writer while I was learning data analytics, my starting salary was in the $50–60K range. I worked closely with people running A/B tests. After gaining some confidence in the first six months I started helping our growth-hacking team and in a year was promoted to the A/B Experimenter role. Meanwhile I moved to study data science and took a few statistics courses. I was eager to learn and test as much as I could, I talk about it in this interview (in Russian only). In a year my manager offered me the Lead Experimenter role which allowed me to grow both financially and professionally. A few years later I got a Data and Technology Director position at a different venture within the same accelerator, I wrote about it previously here.

Bottom-line, if you can, get a job that will allow you to grow from a non-technical to a technical role. It might be easier than you expect.

I think I answered the key questions about how I started. If you have more questions feel free to ask in the comments or in the Facebook group.

List of classes and courses I took:

R Specialization (Coursera, $49/month)

CS50 (Harvard, free) — Great class to gain general idea about programming.

Python & Data Science in community college — Unfortunately the college I went to stopped offering those classes, however I found similar programs here: Python and Data Science.

TeamTreeHouse tutorials

Datacamp tutorials

Statistics (class in Russian, free) — There are various good statistics classes that I can recommend

Data Science Basics (free videos by Prof. Michael Pyrcz, GeostatsGuy) — Michael has videos about Python, Data Science, ML and I found his explanations to be very digestible.

GDG bootcamp (free if you attend all the classes) — There are no current bootcamps, but I am sure there will be more in the future

FastAi (free) — This is an absolutely wonderful ML course that I am currently taking. Highly recommended if you know python.

Machine Learning (Coursera, free)

Additional Resources:

Python and Data Science Learning group on Facebook — We post in this group jobs, classes, and resources. Feel free to join even if you are from the outside of the Bay Area.

AiAi (free for women and minorities) — Check available courses, fill out the contact form on their site, and feel free to ping me if you are interested in being part of the next cohort.

The reason for this post is my own uncertainty at what to do next after finishing a one year program in data science in an adult continuous education school. Somewhere in April 2021, while spending my weekends writing python code for linear regressions, I was haunted by not knowing what I will do next. I already had a data related job, however my job didn’t require the use of data science (we use mostly data collection and some basic analytics), and I knew that if I stopped practicing data science even for a month I would probably forget a lot (since my knowledge wasn’t solid) and it would be hard to get back to it and grow into an expert.

I couldn’t and I didn’t want to quit my job. Thus, I was thinking if I should get a contract gig on the side in data science, enroll in a few more courses on NLP (the part of data science that interested me the most), take part in Kaggle competitions, get an internship, or should I find a proper use case to incorporate data science at my day job?

Talking to my classmates I realized that the majority faced the same problem: they also didn’t know what to do next to keep gaining new skills and acquire domain knowledge while starting to make some money.

I decided to reach out to the community and people I knew and ask them questions with the goal of writing down what are the career options for someone who has finished one year of continuous education in python and data science.

I am deeply grateful to Natalia Muravskaya, Crystal Ramjattan, Tara Jawahar, Tessa Blankenship, Ina Goldberg, Scott Adams, and others for their desire to help, insights, and time.

The format of this post will be questions and answers. It came out quite long, longer than I expected. All questions have been answered by multiple people and I decided to include all the answers to bring more robustness to the story and not throw away valuable perspectives.

TL;DR or grab a cup of coffee and start reading!

Smarts from the Data Scientists

Disclaimer: statements below reflect the opinions of the interviewees and not of any organization.

Question 1: How did you get your first internship or job in data analytics and data science? What did you do in the first 3 years after graduating?

“I have an Applied Math degree, and after I graduated university I started working as a Marketing Data Analyst. A few years ago I was promoted to a Data Scientist in a company where I’m currently working. Learning-by-doing is the best approach for me for the last decade, so I took on-line courses on Udemy and Coursera (Tableau, SQL, Python, R) when a task required a new tool usage.” (Natalia Muravskaya)

“I entered analytics after graduating university. I started in technology consulting and I was very clear with my intention to focus on data and analytics. I’ve continued in the journey transitioning from general analytics to deeper data science and operating in the inbetween to solve analytical problems at large corporations and startups! Throughout my time, I have used mostly SQL and Python which I taught myself via Codecademy and on the job learning. I have always tried to take courses on the side in Deep Learning and Machine Learning mostly through free platforms like Udacity and Coursera.” (Crystal Ramjattan)

“I got my first data internship in the data analytics field by focusing on data work in a previous non-data science internship. I used my first internship opportunity to seek out ways to use data science and improve the processes they were working on, and this experience as well as my academic school work, helped me get my first data science internship. This is my first year out of college, and I am working as a Data Scientist full time.” (Tara Jawahar)

“I was able to break into data analytics while working at a startup. Startup’s are great, you can get experience in different areas because resources are scarce and you do different jobs. I was able to get into BI Development and write reports using SQL in MSSQL and PowerBI. I started working on my education at the same time and was able to move into larger companies and better positions.” (Tessa Blankenship)

“I took a data science position with Nielsen while I was finishing my dissertation for a PhD in sociology. I was really interested in probability sampling methods at the time, which is a really important piece of Nielsen television audience measurement product, so there was a mutual interest. I knew about Nielsen and that the work they were doing aligned with my interests because there were some alumni from my university working there when I took the job.” (Scott Adams)

“While I was self-studying to get into the data field, I looked for jobs and projects that most closely aligned with my goals, to gain experience. For me, working in the nonprofit field, I looked for any role that dealt with data regularly, even if it wasn’t an analysis focused role.

I also created a small portfolio of projects in my free time. For my first data analyst job, I honestly got really lucky — I saw a job post on twitter, and they were looking for a fairly unique background paired with data analyst skills. I was able to showcase my skills through my projects, and the project themes really aligned with the goals of the hiring manager. I’m about 2 months into my role and definitely still learning the ropes. I’ll have to get back to you on what I’m doing in a couple of years!

What helped me find a job:

- Searching for the types of jobs I wanted, and seeing what skills they were looking for, and prioritizing learning those skills.

- Creating a portfolio to showcase the types of projects I wanted to be working on, to demonstrate that I could already do them.

- Doing more impact focused work in the nonprofit world. I think I was able to show in cover letters that I cared about the job and was really interested in the specific role I was applying to. I think that may have helped me get in the door, and my portfolio projects helped demonstrate I had the technical skills, despite the lack of experience.” (anonymous Data Scientist)

Question 2: What was your path into data science? Was it a linear (undergrad/grad school) or a non-linear one (continuous education)?

“I studied Operations Research & Information Engineering which is a combination of computer science, statistics and business. Even so, breaking into data science wasn’t easy. I started as a Data Analyst and then I moved into data science, the first opportunity that I had (after 4 years) which came after a lot of self-learning and portfolio projects! The dots connect looking back and a lot of people I know get into a company, get the analyst jobs done and then move into the data science role. With that said, you can definitely get the role right out of any program if you already have your portfolio of work and can demonstrate advanced data science thinking.” (Crystal Ramjattan)

“My path into data science was linear. I studied Statistics & Machine Learning in undergrad and was able to get a job as a data scientist after graduating. Many of the classes I took in my undergrad helped me learn the skills I needed, but my internships and additional project work are what got me my first job.” (Tara Jawahar)

“I didn’t have any education and was able to get into BI Development at the startup that I worked at. After realizing that I would need an education to get further I started college.” (Tessa Blankenship)

“It was non-linear to the extent that I was not coming from a mathematics, computer science, or physics background like many other data scientists I have met. With that said, I was a quantitative sociologist and was very interested in statistical programming, so that made the transition to a data science career easier. Even with a quantitative background and experience with programming, however, there was still a learning curve going from an academic background to industry. Furthermore, even though I am in a data science career, I am always learning more. Whether it be through books (I love the O’Reilly books with the animals on the covers), online courses, or other avenues, I never want to stop learning. So, in that sense, it is a continual non-linear path I guess…something like a 10th degree polynomial.” (Scott Adams)

“I graduated with a BA in Linguistics, and had a fair amount of undergrad research experience. I went into nonprofit work, while also having the goal of figuring out a more technical/IT focused career path. I self studied for around a year and a half, and attended a free 13 week data science bootcamp-esque program before landing my first analyst role.” (anonymous Data Scientist)

Question 3: Have you participated in side projects, competitions, gigs in data science? Which ones? Did they help you get a job? Did you have a data science mentor? How did you find one? What is one most valuable tool to pay attention to when you are new to data science?

“Yes, I did a few Udacity projects that I could talk about in the interviews. They helped because I could speak end to end about the data science process by doing real world examples. I am also a big fan of mentors. You can find these people at data science meetups or reach out directly to them for a coffee to learn their stories. I also met mentors at work and they helped me bypass technical challenges which really accelerated my learning.” (Crystal Ramjattan)

“I have worked on side projects that are of interest to me, such as contributing to open source packages. I think this is a great way to get involved in work that is currently being done in the field as well as practice some skills. I have a data science mentor that I found through a Python group in the Chicago area (ChiPy), and many of these interest groups exist in cities around the world. A valuable tool to pay attention to when you are new to data science is the work that is happening in the industry, whether it is breakthroughs in algorithms, efficient uses of certain technologies, or innovative ways to use data.” (Tara Jawahar)

“I have some side projects using UpWork where I was able to work on data analytics projects for various business segments. That was a great experience because I was able to grow my skill-set based on what the client needed, learn how to deploy code in GitHub and learn how to work in cloud based systems like Google Cloud Platform and Amazon Web Services. The future of data science is in the cloud.” (Tessa Blankenship)

“I have been involved in teaching SQL, Python, and data visualization for a great startup in Chicago called Promotable that is focused on data education. I also publish articles and tutorials in places such as Towards Data Science on Medium. I really enjoy sharing the lessons I have learned in my journey in the hope that those lessons can help others break into data careers. I have found that teaching and writing has been crucial in helping me become a better data scientist. Teaching and writing both require you to articulate content in a way that is accessible to those who are unfamiliar with said content. In the process of putting together content for a course or article a lot of my blindspots become apparent and I often have to go back and re-learn things I thought I knew pretty well to explain them in a coherent manner to others.” (Scott Adams)

“I haven’t participated in any side gigs or projects that were paid, but I did do my own side projects on topics I was interested in. I definitely think being able to showcase and talk through my approach to projects really helped me get my position. I had a very helpful mentor when I was first trying to figure out if I wanted to go into software engineering or data analytics, and I reached out to several other people over twitter throughout the time I was self-studying.

As far as the most valuable tool — I think it really depends on what sort of role you’re looking for. I’d check job descriptions and see what they are asking for- though a solid programming foundation in Python or R is likely a must for many roles.” (anonymous Data Scientist)

Smarts from DS Team Leads and Hiring Managers

Question 1: What are you looking for in an entry level candidate? What base skills should they have? Could you expand and tell what skills your most junior data people have?

“Six month ago we hired two entry-level Data Scientists to the team. We used the following approach: after the first 30-min screening interview, we asked a candidate to do a homework assignment, and if the result met our criteria, the candidate went to the next interview round. A successful candidate had to demonstrate programming, statistics, machine learning, data analytics, data visualization, and critical thinking skills.” (Natalia Muravskaya)

“In entry-level candidates, we tend to screen candidates for a portfolio project they can speak to that demonstrates key skills: SQL, storytelling with data, basic visualization skills, and clarity in communication. For example, we just hired someone from university who majored in business and learned most of her skills on the job. Another junior level hire has a liberal arts background. They are both brilliant analysts and bring a unique perspective to the field. I think what stood out in the interview process were their general problem solving abilities and their willingness to jump in and learn on the job.” (Crystal Ramjattan)

“We do hire entry level candidates and look at their ability to learn new technologies. Being able to show that you are a constant and excited student would be what we look for in these candidates.” (Tessa Blankenship)

“When we hire freshly minted data scientists, we are looking for a rounded balance of math and statistics, coding and domain knowledge. We are also seeking candidates with sufficient communication skills, able to explain their work to both technical and non technical audiences. MOOC and Bootcamp graduates may find it challenging to compare favorably with the graduates of dedicated BA/MA programs, focused on building strong academic foundations and real world data science experience during their studies. However, having a strong domain knowledge can be a huge plus in certain industries. It is also extremely helpful when data science candidates are able to connect their domain knowledge with newly developed technical skills, and translate these into suggestions of models and analyses. It is even better when they are able to demonstrate these practical applications in their github portfolios”. (Ina Goldberg)

“It depends on the specific role. With that said, there are a few common characteristics I look for. With respect to the non-technical side, I look for two things in particular. One, an internal motivation to learn. The data and technology space is always changing, so it is really important to find people who want to learn new things. Second, I look for strong communication skills. Data professions involve a fair amount of collaboration with individuals across different roles, so knowing how to clearly articulate your ideas, needs, and challenges to various technical and non-technical audiences is crucial for moving a project forward. On the technical side, some familiarity with Python and/or R is helpful, but I also want to emphasize the value of knowing SQL for entry-level candidates. While it is not as glamorous as running a neural network, SQL is really foundational because in most data careers, be it data science, analytics, or data engineering, much of the data you need is probably going to be stored in a relational database that will require the use of SQL.” (Scott Adams)

Question 2: Where do you post your job openings? How do you find the people you want for internships and junior roles? Do you focus onthe college degree or the skills the person has?

“One of our recently hired Associate Data Scientists was a Research Physicist before, and he finished the Springboard Data Science course to convert into the data science field. The other candidate had a Business Analyst experience for 5 years before being hired as an Associate Data Scientist into the team. She used Udemy and other online courses to acquire skills needed for data science work. Among necessary professional skills they demonstrated during the interview process, we also checked: how well he/she could fit the team, their communication skills, candidate’s career goals, readiness to learn, and curiosity.” (Natalia Muravskaya)

“Job postings are on our LinkedIn and career website. If you apply there, it’s a game of chance. Just use the postings to know someone is hiring but your next step is to get noticed: stand out from the crowd.

- Reach out to someone in the company. Pay attention to hiring managers that have “HIRING” in their LinkedIn job title.

- Most people I saw hired were via a recruiter or referred into the company. So network with people that already have the job you want or find recruiters!

- I think most hiring managers patterns match: they want to ensure you have some examples of doing the job: have technical skills in Python/R, SQL, data visualization, data processing, analyzing key metrics, etc. Once you have some clear examples, it’s about communicating that in the interview process.” (Crystal Ramjattan)

“The main areas that I would look for would be projects that they had completed and if they were able to explain the fundamentals. We haven’t hired any new junior staff since I have been in my current position.” (Tessa Blankenship)

“We hire for entry level positions from all paths of life. How do we find the right candidates? We advertise in community college boards, bootcamps and share information in various data science communities. Some candidates come through word of mouth: recommendations of other data scientists or leaders in the industry. My advice to bootcamp graduates: start from data science internships in the industry you know well and have worked in before. Knowledge of the industry, understanding of use cases and available data is a huge plus. I would be more willing to hire a bootcamp student for an internship if she has a deep knowledge of the domain, while able to demonstrate the technical skills. Sometimes, transitioning from the IT background can offer an extra leverage in terms of knowledge of underlying data and systems.” (Ina Goldberg)

“In my experience, companies looking to hire data professionals will have some kind of recruiting footprint on sites like LinkedIn or Indeed. These are great places to look for a career. On LinkedIn in particular, it is helpful to adjust your settings to let recruiters know that you are actively looking for a new role.” (Scott Adams)

Question 3: Did you look at side gigs that your entry applicants did? Which gigs impressed you in the people you interviewed? Did you ask technical questions in the interview? Did you ask entry level applicants to write code or solve data problems? What is the best way to get trained to solve these problems?

“We asked candidates to build a model that implements a machine learning algorithm on the received data (it was a homework assignment). To get work done, a candidate had to:

1) Perform exploratory data analysis

2) Choose method(s), select features, train and execute selected model(s), evaluate the model(s) performance

3) Prepare a slide deck to present results

4) Provide model(s) code and explanations of findings and methods used.

Based on what they sent us, we selected candidates for the next round of interview, where they presented their findings and were asked about the details.” (Natalia Muravskaya)

“In our company, we do a presentation interview where we give candidates data and ask them to bring us a unique perspective to a problem. They present it to us so we can see end to end how they solve problems and how they communicate their findings. You can use Datacamp / Kaggle to get some basic practice. It should be a good way to learn key skills we like to see in the interviews.” (Crystal Ramjattan)

“Yes. If an applicant can show some of their ability to learn and adapt to the needs of different business segments that is impressive. We do ask them questions related to code and have a code challenge that they need to take. This involves different flavors of SQL and a Python challenge. We really focus on the logic behind how they solved the problem, so it is less about being perfect or 100% correct and more about the thought process.” (Tessa Blankenship)

“I will say this. If I come across an application I am going to first make sure that they meet the minimum requirements for the role. Then, I consider any relevant item(s) demonstrating that the applicant has something to offer to a team. Education, full-time work, part-time work, side-gigs, passion projects hosted on GitHub…you name it. With that said, I also want to emphasize that it is important for an applicant to showcase what really makes them unique and to be concise in listing their skills and accomplishments.” (Scott Adams)

Advice from Folks in Data Science

From your personal experience, what advice would you give those graduating from a one year data science certificate program on how to cross the chasm of “no experience” and get their first internship, side gig or entry level job?

“Due to high demand for Data Analytics and Data Science professionals, there is a good chance for an entry-level specialist to get a Data Analyst job first. Thus, start applying to Data Analytics openings, particularly, there are a lot of contract roles on the job market now. I consider this as a real opportunity to gain minimal experience and start building a portfolio of projects that you can demonstrate to your potential employers. So, with time and more advanced tasks gained, you will naturally transition into a Data Science field.” (Natalia Muravskaya)

“The sky’s the limit. Focus on what you bring to the table. Have your portfolio of 2–3 projects ready and talk about what excites you most about the projects & the field. Apply — always throw your hat in the ring. You deserve this and you can get the opportunity but you have to apply. Build your community. Attend meetups. Work on projects for yourself, with others or for non-profits. Write about your perspective on Medium or LinkedIn. Offer value to the world and then let people know that you’re looking to continue growing your skills. Most data scientists are self-taught by reading blogs / books and build experience through real world examples. You can do this! I hope that helps but I know you can do it — you just have to believe it too!” (Crystal Ramjattan)

“From my personal experience, I would encourage women to engage in data science-related communities to network and also improve their technical skills. This could include contributing to open source work, finding a Meetup group or similar community in the area, or reaching out to someone in academia or industry to learn more about what is happening in the field.” (Tara Jawahar)

“Always be learning. The side gig works, because you get experience and you show your ability to be flexible and learn new business segments. The imposter syndrome is real. Remember that everyone has to look up code, no one knows it off the top of their head, so don’t feel that you are not good enough at coding or not good enough at analysis, or whatever it is. If you keep an open mind to learning and are not afraid to ask the “dumb” questions, you will find that businesses are looking for someone like you. No one likes to work with a “know it all”. (Tessa Blankenship)

“Show you can do the work through a portfolio! Find a dataset and a problem, and figure out a way to gain insights from the data, and then present it in some form (I like pretty dashboards) that is easily understandable. I suggest approaching it as a hypothetical business problem; decide on an audience and suggest actions to take based on the data (if applicable). If an interviewer asks you in an interview about a past problem you’ve solved, have a dashboard on hand that you can bring up and walk them through.

I also found it very helpful to lean on my other work experience — it’s one thing to be able to code, but it’s another to be able to address the actual problems that you tackle in a data science or analytics role. Sometimes having specialist knowledge or past experience in certain areas can really help you make insightful conclusions (and land a role).

The job hunt is pretty painful. I did ignore most “years of experience” requirements; if I had some of the skills they were asking for, and the role sounded like a close match to what I wanted to do, I applied.

And above all else — you got this!! If you made it through a bootcamp you’re already doing amazing! (And, one last thing; if you’re in a position to do so, negotiate the heck out of your salary)” (anonymous Data Scientist)

“For a junior Data Science candidate, it is also helpful to be able to demonstrate aptitude for following a hypothesis driven process, which can be tedious and iterative at times. After all, a large portion of effort in Data Science is dedicated to data prep, and starting from that foundation is invaluable for early career experience. Junior data scientist roles might be focused on ETL and data collection, data labeling and parsing, feature development and hypothesis validation. As the fields of Machine Learning and traditional Statistical Modeling continue to converge, these activities are becoming increasingly important for various branches of data science.

While knowledge of advanced methodologies is important, it often takes a few years to reach a decision maker level when it comes to method selection. Ability to do it successfully, is usually grounded in the on-the-job training. Like in other disciplines, acquiring the needed 10,000 experience hours is a very valid concept in data science. My advice to early career data scientists: do not underestimate the amount of real skill and knowledge acquisition available to data science interns, collaborating along with the experienced team members.” (Ina Goldberg)

“It is helpful to recognize the value of networking. Keep in mind that professional social ties can be really important in expanding your awareness of new career opportunities. Building networks, even virtual networks through LinkedIn or other similar reputable online social networks, can expose you to opportunities you may not have thought of otherwise. Through network-building, you also increase your chances of finding a mentor that can provide you with advice and other forms of support along the way.” (Scott Adams)

Apart from doing interviews I got a chance to discuss this question with Gregory Renard and here are a few things that stuck with me from this conversation. One, you have to believe in yourself and do things that others (and sometimes even you) think impossible. Greg showed me a small figure of a magician he has on his desk, because data science should feel like magic to others. Second, you need to find a mentor. According to Greg it is crucial to have one or more mentors. How to find one? Reach out to all the people whose work you admire and ask them to look at your code and to mentor you. You will certainly have very few returns for many reasons including the fact that their agenda might already be very busy, but some will get back. At the end you will find the right mentor. Third, find a peer group or a buddy with whom you can talk daily about the data problems you have, challenges, code, etc. Communication and talking the ideas out loud with someone is crucial to developing good ideas. Fourth, start by solving your own data problems, notice what things you do in your digital world that you can automate. Study your own behaviour when solving these problems and translate it to code. Fifth, learn more about data and information overall. We live in a world where information is everywhere but almost nobody is extracting and using it. Just think about this and all you can do with it! Here is a book that Greg recommended reading. And last but not least, have strong ethics. If you work with data you will impact the lives of other people, possibly millions of people. Make sure this impact is positive.

For those who’ve made it till the end of the post, I hope your coffee is still warm or at least you are ok with its temperature. I often think that when I will be hiring for data science roles one of the questions I will ask is “How do you feel about drinking cold coffee all the time?” And if the candidate is fine with it, she(he) is a true hiring material. I think that 50% of success of being a good data geek lies in having flexibility with your coffee temperature.

As a conclusion of this long post I want to write my personal insights and my plan (that I now have!).

First of all, through conversations and projects I found the area I want to focus on for the next six months: data pipelines, and specifically data pipelines for linguistic data. Second, I will need to find a mentor who can give me feedback on my code and thinking. Third, I will find a buddy to discuss my code, share resources, and collaborate with. Fourth, I have made myself a list of online classes I want to take to hone the skills in data pipelines and NLP. Fifth, I decided to drop one of my side marketing consultant gigs and find a small data science gig instead (I will drop the smallest gig and there will be no big impact on my income, actually my plan is to make 20% more). Sixth, I will mentor one or two people who are just switching career paths into data (I’ve been there 3 years ago). And seventh, I will try to be helpful in my jobs using what I can from my arsenal.

And here is the most important bit, most of the data scientists I talked to ended our talk with “Believe in yourself!” I think they are onto something.

The goal of writing this blog post is to help others, and especially women in tech, get their dream job position. I have mentored a few people at my current job using my own experience and I decided that writing it down might help another handful of people.

This is how it went:

I had my annual review in March 2020. My manager asked me how I wanted to grow within the company and this made me think of where I really wanted to be and what I wanted to do day in and day out. At that point I was a lead experimenter and managed other experimenters, and my job consisted mostly of running lean experiments in the chatbot space, analyzing data from those experiments, reporting results, iterating, and scaling successful models.

After my annual review and evaluation meeting I opened a new Google Doc and wrote my dream title “Technical and Data Director”. I set myself a timeframe of one year to get there.

Here is what I did next:

1. I went on Glassdoor.com and I read through fifty or sixty job descriptions for CTO, Data Directors, Data Managers, Technical Managers, Technical Directors, etc in the same space.
2. I copied and pasted all the qualifications that I found often in those job descriptions in a Google Sheet and that could be applicable to the company I worked at. I ended up with quite a long list of skills and experiences.
3. From our HR department I obtained my most recent job description. I made a list of skills and experiences I had based on the job description and based on what I actually did for the past two years at my job.
4. I compared both lists. The list of the CTO type skills and the list of my current skills as the Manager and Lead Experimenter. Majority of skills and experiences overlapped. I created a third list, a list of the skills and experiences I needed to gain. I looked at this list and took away those that I deemed non-crucial. I ended up with three “must-have” skills that I knew I lacked: Data Science, Strategic Management, and Data Systems Architecture.
5. I asked for a meeting with my manager and showed her my spreadsheets and my findings. This way I got her to buy-in. I explained to her that I would be more helpful to the company in a Tech Director role and that here was the list of skills I needed to gain to get there. I asked for her opinion and insights. She told me she needed some time to think about it.
6. We met again after two weeks and she told me that she considered my path and plan reasonable, however after she ran it by the company’s HR and HQ she was told that to aspire to the director’s role I would also need to be able to prove a financial model for a business (or a new branch of a business). I told her that I needed some time to think about it and that I would get back to her.
7. I went forward and looked at every director at our company. While majority of the directors had proven a financial model for a business and run their startups (in the company I work a director role is a C-level role), there were two directors who were not running a startup and have never proven a financial model (they were the Director of Finance and Operations and the Director of Venture Strategy). I asked for another meeting with my manager and showed her these examples. Long story short, after two more meetings HR agreed that proving a financial model shouldn’t be a “must-have” for a director role. I would like to add that I like working on financial modeling and I think it is vital to understand how financial sustainability works and to build (or be part of a team that builds) the financial model. However I didn’t want it to be a requirement for the role I was aiming at because finding a proven financial model for a business can take years.
8. By the time we finished our “negotiations” it was already June and I had about eight months to gain the three skills I needed: Data Science, Strategic Management, and Data Systems Architecture. I reached out to other CTOs and CEOs and asked for recommendations of a coach to learn strategy and strategic management. In a few weeks I got introduced through a friend to an amazing coach, Jeff Malone, and we started our work together in September. Jeff helped me create a strategic method for experiments, develop a management style, and set the team I was managing for growth while setting boundaries and expectations. Since then “The Strategic Method” he taught me (and that I wrote down diligently) spread out into other divisions of the company and people started reaching out to me to teach them and let them use it (which I did gladly).
9. In July I signed up for a Data Science course in a community college. Since the beginning of COVID all education has moved online and I was able to find a great and affordable course with a real teacher. Spending my nights in class and weekends doing homework, I quickly gained some skills (not of a data scientist, I didn’t aim at that) to do data explorations and orient myself in predictive models.
10. Learning data systems architecture turned out to be a bit more challenging. There was no concise course to take, there was a variety of platforms and solutions, and our work with outside partners, chatbots, and Facebook had its own requirements. I ended up digging information through Twitter, reading tons of articles, and looking at many different platforms and tools. I took a few classes on Datacamp about databases and systems. I reached out to our internal IT department to read the documentation on our current data structures, asked them what other similar companies are using, and engaged in a series of learning sessions with our IT team. I reached out a lot and many times I felt terrible taking up people’s time, but this was the only way to learn.
11. I made sure I kept my manager posted on my progress. Each check-in I included a short update on my professional development and where I was with each of the skills.

What happened next was a bit more exciting than I expected. In February 2020 (before my annual review and before I started this whole process) I was asked to jump on a side project as a tech support. I built a chatbot architecture for a small project within the company. We made it possible for organizers to go digital and use chatbots and automation to gain constituents, virality, and generate actions. I was first expected to spend five hours per week on this project, then ten hours per week, then twenty hours per week. By mid July I was working over sixty hours per week. Twenty of them on my “full-time project” squeezing all my work in twenty hours and rigorously skipping meetings and fourty hours on this side gig. The side gig gained traction and grew. In August we hired our first two employees to help build our tech. In October the side gig got funding for a two year runway. In November the CEO of the project asked me to step up in a CTO-type role as a Technical and Data Director. I was excited about the offer and after a brief back and forth accepted the position transitioning into the new role on December 15th 2020.

Since then I have been approached by my previous team with questions about the growth and how they could get the positions they wanted. I shared my process with them. I am currently coaching a part of my old team on professional growth. I am also coaching my new team how to get to their goals and dream jobs within or outside of our comapny. The process is very simple and can be summarized as follows:

1. Write down the job you want to do in two, three or five years.
2. Find twenty plus job descriptions that fit your dream job.
3. Make a list of your dream skills and your current skills.
4. Create a list of skills you need to gain for the dream job. Prioritize.
5. Keep you manager involved and helping you.
6. Make sure HR and company management think your plan is reasonable.
7. Set your plan in motion: sign up for classes, read, get coaching, volunteer, do side gigs, basically do whatever it takes to gain these dream skills.
8. Keep your manager updated on your progress and adjust where needed.

I think that opportunity shows up when you put the work, not before that. I was extremely lucky that I got asked to help on the side gig. However, it was my decision to accept it, knowing that it might take more time and I will have to work long hours. It was luck that our side gig grew that fast, however it was us that did the job and built all the tech. Without the builds luck wouldn’t show up. I was lucky to get offered the job I wanted before expected and in a different scenario than I envisioned it. However, it was not a gift. I remember that I wasn’t even happy when I got asked to be the CTO. I worked so hard to get there and was so tired when it happened that I just felt that it was a natural thing to occur.

A month later, around the New Year holidays, I told my kids and my family and friends about the promotion. And that was the first time it dawned on me that I’ve reached my goal. And I was happy that the process worked and started sharing it with those who asked. I even shared it with our HR department in hopes that it will become a “standard” process and help others. For the same reason I am writing it here. I think it works if put in action.

I have to add that the only thing that wasn’t hard work and was pure luck was having a good manager who was supportive of my growth. I don’t think this process would have worked that well otherwise.

To increase your click-rate from your chatbot:

1. Name the day of the week or add “time location”: Starting your message with “This Thursday learn how to cook a healthy dinner” makes people click more than when you just say “Learn how to cook a healthy dinner”. Here are a few more examples that worked for us: “End the week well by learning the lyrics of this popular song” (sent on Friday), “Start your week-end strong by learning how to talk about road trips” (sent on Saturday) and “Happy Monday! Today we are going to learn how to write a good resume”.
2. Use emoticons: adding 2 or 3 emoticons in the right place will make your message look fun and easygoing. The messages such as shown below scored quite high on click-through rates. In our case I took emoticons from FB messenger, since this is the channel where our users receive it.

3. Go for text+image+text structure rather than list or gallery cards: This might be something relevant to Chatfuel users only. Our click-through rates improved aprox. 1% when we used “text+image+text w/buttons” structure over the list card. The only issue you might face is a small time gap between receiving the first text and the following image and text. Chatfuel is aware of the problem and working on fixing it.

4. Use female smiling face in the image: When it makes sense use an image of female smiling face, it works better than any other image. People are very visual and using the “right” image makes a big difference. I would not recomend using runway models’ images, however a “real woman” smiling face worked in our case.

5. Include a trigger in your broadcast: If you want your users to click to your site or app pick their interest with a question. What worked well in our case is posting a picture of a word and asking users if they knew what this word meant. Below is our broadcast in Spanish and the first text block reads “Masha, do you know what this word means?”.

6. Use fear aspect in your copy: Fear of missing out is still a huge lever that propels people into action. Phrases such as “Don’t stay behind your classmates” and “Don’t lose your chance to improve your English today” worked well in our case.

Two other things that you can use as add-ons (i.e. we have not tested them in isolation and I can’t say if they bring any significant improvments) are reminding users what time commitment you are asking from them and highlighting (low/non existant) cost of your service. In our case we remind users that learning English with our chatbot is free and only takes 10 minutes per day.

Feel free to comment with what works for you!

Since I have started writing about chatbots and run a few bot building workshops I have been asked numerous times on what bot building platforms I can recommend. I put together a list of chatbots building platforms. I will try to keep it updated as new platforms appear. If you have used a platform that is not in the list feel free to leave a comment below.

Alan

Keywords: Collaborative technology platform, integrational speed, easy to use, human communication, LATAM, Spanish

API.ai

Keywords: NLU, conversational agents, developer focused

BotAmp.com

Keywords: Marketing automation, small businesses, eCommerce

Bottr.me

Keywords: One-click bot, avatar bot, for events, customer support, influencers and busy professionals who want a personal bot

ChatClub.me

Keywords: Chatbots for Facebook Messenger, for artists, sport stars and celebrities, marketing focused, push notifications, no coding required

Chatfuel.com

Keywords: What you see is what you get, CRM console, completely free

ChatterOn.io

Keywords: Build a bot under 5 minutes, pre-built chatbots, end-to-end, multiple platforms integration, ML, UX and AI assistance

Chattigo

Keywords: Customer service bots, Spanish language, LATAM, cross platform, human and AI, CRM integration

Conversable

Keywords: Enterprise solutions, any messaging or voice channel, ecommerce & brand building via messaging, conversational intelligence flows

Converse.ai

Keywords: Build without writing code, workflow tools, drag and drop designer, human in the loop

FlowXO

Keywords: Flexible, customizable, user support

Gupshup.io

Keywords: Toolkit for bots, entire bot lifecycle, bot under 2 minutes, for non-developers

Howdy.ai

Keywords: Startups, Slack, Developers, Code

Init.ai

Keywords: NLP, machine learning, monitoring conversations

Kitt.ai

Keywords: Chatflow, NLU, bot-to-bot communication, drag and drop, Alexa integration

Kore

Keywords: Ai-rich conversation interfaces, enterprise-grade, end-to-end, rapid deployment

Manychat

Keywords: Facebook bots, no coding required, free, build a bot in 2 minutes, autoposting

Msg.ai

Keywords: Conversational ecommerce, cross-platform, multi-channel, enterprise solutions, A/B testing

Meya.ai

Keywords: Human/bot hybrid, developer focused, bot flow abstraction

MindMeld

Keywords: Voice & chat assistants, deep-domain conversational UI, undersatands questions and provides responses, end-to-end support from prototype to production

Microsoft Bot Framework

Keywords: Open source, cognitive services, cross platform, server less

Motion.ai

Keywords: Botstore, easy bot building, complex bot flows, small businesses, no coding required

Octane.ai

Keywords: Content creators, marketing tools, fast bot creation, no coding required

Pandorabots

Keywords: Multilingual, speech interfaces, free, provides education on bot building, AIaaS (Artificial Inteligence as a Service), customizable AIML, 225K+ developers community

PullString.com

Keywords: PullString Author, personality for bots, human touch in bot conversation, Alexa skill

Recast.ai

Keywords: Sharing, simplicity, luxury brands, public services, no coding required

Reply.ai

Keywords: Customer service bots, integration into any customer service engine, built-in CRM

Sequel

Keywords: Conversational experiences with personality, interactive entertainment, drag and drop, perfect for media and entertainment industry

Smooch

Keywords: Operationalizing messaging experiences, unified API, one-click integration, multi-channel, developer support

Tars

Keywords: Convert web-forms into chatbots, works with any language, export user conversations into excel, no coding or tech skills required

Turbo.ai

Keywords: Real time bot building, instantly, no computer skills required, perfect for restaurants, hotels, small businesses, free

Ubisend.com

Keywords: Conversational marketing platform, no coding or tech skills required, for marketers, free for charities and NGOs

Watson Developer Cloud

Keywords: Free using Bluemix, NLP, multi-channel, tutorials & demos on the site, free & paid plans

Webio

Keywords: Customer service intelligent agents, all channels, live agent + Webio Ai agent, instant dashboards and reports

Wit.ai

Keywords: Voice recognition, free, developers’ community oriented

You can read in detail about some of the bot building platforms here

Opensource tools

ChatScript the “original” Bruce Wilcox’s language/engine; still under active development

SuperScript made in JS Node, is a sort of simplified ChatScript

RiveScript a simple scripting language that inherit some AIML and ChatScript idioms

You can find more information on opensource tools in this blog post by Giorgio Robino. I appreciate his help with this list.

Thank you for reading!

Masha works with bots & NLP | Founder & CEO at Bluewordai.com, Arithebot.com and FiveperFive.com | You can find her on Twitter and Linkedin

Chatbots Building Platforms List was originally published in Chatbot.com Blog on Medium, where people are continuing the conversation by highlighting and responding to this story.

(notes from a bot user)

Majority of the bots will not understand your language. In the past we have tried a great deal of bots from finding a place to eat to trying to order some Pokémon cards and after the first greeting the conversations went to “I do not have the answer for this” or at best “Let me connect you to a human assistant to answer your question”. Rarely bot understood a normal query done by any of us, humans.

As a consequence, fired by the desire to be the first to use this new technology, we ended up talking to bots in pigeon English. Personally, if I am 100% sure there are no human on the other end I skip greetings and my queries are very straightforward and deprived of complexity, “order pizza”, “find a flower shop”, “buy tickets”. Mostly “verb + noun” type of structures. And the bots are responding much better to these simple phrases. A bit of humor here: Instead of us training the bots to understand human language, the bots had trained us to reduce our rich English to a few practical and laconic phrases.

Still the problem remains the same: bots are our new toys and there is no mutual understanding. Not yet. One of the ways some companies are solving it is by creating “buttoned-up” solutions and at the same time keeping the option of free text. Rémi Meunier, COO of Recast.ai, said that while developers prefer using natural language majority of corporate clients prefer buttons. “Buttons are powerful in terms of user experience and people respond better to buttons than to having to type their own entry. One of the reasons for this “buttoned-up” bot culture could be that many of the companies do not know what they want from the bot and as a consequence expect it to behave as the website does,” says Rémi Meunier. You click here, we take you there. This is quite far from a real human conversation. Actually this is taking us two steps back from the AI-forward future.

Still we want these companies to build their bots. Buttons or no buttons. They are building the pillars for all the conversational interfaces, NLP and deep learning to come. They also get people used to the idea of bots and talking to a machine instead of a human being. The most important thing is that we are moving forward by building bots.

Here are three valuable things about bots:

-Bots allow two-way communication (value: you can build a relationship with a bot and have a meaningful dialog).

-Bots can solve problems and complete certain finite tasks much faster than a human being (value: saves time).

-Bots have no judgment (and we as users know it). This means we can ask bots any questions that they can answer without the fear of appearing stupid or weird. (value: it is like talking to your toaster. Just kidding. The real value is we feel safe talking to a bot).

These three things combined together can open doors to a new way of discovering the world (read here: learning) and gradually change our learning behavior. It is like having a personal assistant and a teacher at the same time that never thinks that you are stupid or have poor memory. It actually accepts you as you are and adapts to your learning practices and pace.

One thing that bots will need to learn is appreciation. As humans we want to feel appreciated. All learning theories show that highlighting what we do well makes us improve faster. Imagine a bot that always reminds you of your wins, never scolds you for forgetting your past experiences and does not make a long face when you ask about something you are expected to know from long ago.

What could you not become?

Pause for a moment. Really.

What could you not learn in these conditions? You could learn anything you want.

At the point we are right now it will take at least a few years to get to the day when bots can understand our language in its complexity and keep the conversation engaging for long periods of time.

The path is right though. Enjoy zigzagging into the future!

Thank you for reading!

Masha works with bots & NLP | Founder & CEO at Bluewordai.com, Arithebot.com and FiveperFive.com | You can find her on Twitter and Linkedin

Bots do not understand you: still they are the future. was originally published in BotPublication on Medium, where people are continuing the conversation by highlighting and responding to this story.