Talk on Data

Science
Varshit Dubey

DS & AI Consultant
Boston Consulting Group
About me
Researcher by heart Business strategist by mind

Mantra: "If you can't explain it, you

don't know it."
Stats for breakfast, code for lunch, data for dinner,
I believe in the power of clear communication and and business strategy for dessert, discuss over coffee
simplifying complex data concepts.

Proudest moments:
 Watching my parent’s tears of pride as I graduated 2nd in IIM Calcutta PGDBA, receiving an award

from Mr. Uday Kotak.

 Playing a key role in ensuring 100% placement success as part of the college placement team.
 Mentoring 50+ students on crafting standout CVs and nailing interviews
Agenda
1 My Journey into AI 2 Career options in Data Science & Skills
required
I'll share my personal journey, from early interests to my
current role at BCG. We'll explore various roles, required skills, and the current
landscape of the data science field.

3 My Journey @ BCG as a DS Consultant 4 Thoughts on how to be Industry Ready

in College
I'll highlight my experiences and insights from working as a
We'll discuss practical tips and strategies to prepare for a
data science consultant at BCG.
successful career in data science.

5 Common mistakes to avoid 6 Q&A

What comes to your mind when you
hear about AI and machine learning?
What I think of AI ?
My Journey into AI

2017: Electronics? Not for me! 1

Discovering my passion for mathematics and exploring data science
through online courses.
2 2018-19: ML Research @ IIT Bombay & Canada!!
Internships and research projects led to the publication of two ML
papers, fueling my enthusiasm.
2019-20: Master's or…? 3
Weighing different options, seeking a blend of technical and business
skills, leading to a decision to pursue PGDBA.
4 2020-21: Confusion Mode On!
Deciding to combine management with data science, preparing for CAT
and PGDBA, embracing a multifaceted approach.
2021: Top Admit! 5
Gaining admission into IIM Calcutta's PGDBA, a significant step in my
journey, marking a turning point.
6 2022: Consulting Life Begins
Internship at BCG, gaining valuable real-world experience and exposure
to consulting methodologies.
2023: Full-time DS & AI Consultant 7
Embracing a full-time role at BCG, ready to make a difference and
leverage my skills to solve complex problems.
WHY AI & DATA SCIENCE
• Are you someone who
• Numbers speak for themselves: >200 Billion USD investment in
AI companies, 2x-3x increase in job postings related to data
science.
• There is no industry right now where data science is not used!!
• Currently, there is a clear shortage of specific skills required to
build scalable and reliable AI products.

There will be more data than we can possibly analyze, unless of course we
have skilled people like you 
Types of roles in data science

Non Technical Mix Technical

Data Analyst Data Strategist Machine Learning Engineer

Business Intelligence Analyst Data Science Consultant Data Engineer

Business Analyst Data Architect MLOps Engineer

Big Data Engineer

AI/ML Scientist/Researcher

Systems AI Engineer/Edge AI
Lets understand each role with an
example
Skills for technical roles: MLE, MLOps
& Data engineer

Be very good in Python, Learn Cloud ETL pipelines Docker, Kubernetes,

SQL, and Big Data tools (Start with AWS) MLFlow, Airflow,
Jenkins, Terraform

Yes, because apparently Because who doesn’t love Become the magician who Don’t worry if these sound
you’re expected to know burning through cloud can move data around, like foreign languages—
every single query credits while trying to and when it inevitably fails just learn them all. You’ll
optimization trick in the figure out what "Lambda" at 3 AM, you’ll be the only spend your days
universe. If not, well… and "S3 bucket" actually one who can fix it. Isn’t containerizing models,
good luck! mean? After AWS, make that fun? orchestrating workflows,
sure you learn GCP and and automating
Azure too everything.😊
Skills for Technical DS Roles II
1 Publish Research Papers
Easy, right? Just crack open a few academic journals, get into the depths of algorithms
like it’s bedtime reading, and voila—you’ve got yourself a research paper. (Or not.) 📄
🤯

2 Algorithms and Math

If math isn’t your bestie, don’t even think about this path. Otherwise, good luck
explaining all those equations to your friends. 😬

3 Courses & Certifications:

No magic course will make you a ML researcher.

4 Try to apply for research fellowships at IITs and international colleges

Because nothing gets your foot in the door like cold-emailing professors with 0
responses for months. When in doubt, keep spamming inboxes. Professors love
that. 🙃
Skills for Non-Technical DS Roles
• Be good at Progg. & Algos: Sure, it's "non-technical," but you'll still need Python, SQL, and maybe five other languages. Easy, right? 😅

• MLOps pipelines: No need to be an expert. Know what tools to use, when, and roll up your sleeves when things get dirty. Project Pro & Applied
AI can “help” you. Jenkins, Docker, Kubernetes—learn them all. 🙃

• Business acumen & Comms: Turn vague business problems into data problems and explain AI to people who think it means robots taking over.
Good luck with that. 👔 📊

• Flexibility: Tools will come and go like fashion trends, but your ability to keep learning? That’s always in style.

• Any Course: Nope, no magic wand here either. But hey, there are some solid programs like PGDBA at IIM Calcutta, PGPBA at IIM Bangalore, and
other MBA in Business Analytics courses if you're feeling fancy.

The best data scientists aren’t the top software engineers, business experts, or statisticians—they're the ones with more
business sense than the engineers, better communication skills than the statisticians, and more technical expertise than
the business folks.
How AN Enterprise Level data science project
looks like?
 My Journey at BCG

2 Countries, 100+ flights,

Countless hotels, mentoring
50+ students – UNFORGETTABLE
MOMENTS!
My Experience: Hectic but Rewarding
Why BCG? Highlights and Key Realization?
Its all about diversity—I wanted more than Takeaways  Simpler models often work just as well as those
just coding. It’s not just about building models complex RecSys models. Oh, and GenAI? Game-
but generating actual $$ from them. (No  Worked for a Steel, Oil & Gas, Consumer
changer for sure, but let’s be honest... extracting
offense, coders! I’m a former C++ dev and Goods, and Telco clients.
$$ from it is still no walk in the park!
loved it 💻).  Built from complex recommendation
 Explainability is everything. If you can’t explain
It’s a client-facing role where I’m a part data systems to Agentic AI to Computer Vision to
why your model behaves the way it does, it’s
scientist, part business strategist, with a good ol’ regression models.
useless—no matter how high the accuracy is.
sprinkle of management spice!
Million dollar question: how to be industry ready?
Master the Basics
Get comfy with stats, algos and coding! 🧠 Start with the ISLR book, your ML bible!

Projects
Build your portfolio by working on end-to-end projects. Platforms like Applied AI or Project Pro are great to get hands-on experience.💼 🔍
Contribute to open source projects on Github. Learn MLOps once you’ve worked on some ML/DL projects.

Compete and Learn

Flex your skills on Unstop and Kaggle competitions! 🏅 🔥 Nothing like some friendly rivalry to sharpen your mind!

Internships are Gold:

Try to get an internship. Hunt down for research fellowships like FOSSEE at IIT Bombay, Mitacs in Canada, DAAD in Germany,
USC Viterbi in the USA, or NTU in Singapore. 🌍 📈

Research time
Got an idea? Publish ML research papers 📝—leave your mark on the AI world!

Speak up and network

Boost your communication skills . They pay dividends in the long run! 💰
My Personal and Professional
Projects
Common Mistakes to Avoid

For the love of God,

Jesus, and all
things holy—please
don’t do this!
Common Mistakes
to Avoid

I started out thinking data

science was all about
algorithms and research,
but quickly learned that
communication and
storytelling with data are
just as important.
High-Demand DS Roles in
the Future
1 MLOps Engineers
Increasingly sought after for integrating machine learning into operations. Last
decade was all about experimenting with AI models. This decade will be about
scaling and operationalizing the models to production.

2 Data Science Consultants

High demand for those with business acumen to identify use cases and drive
value.

3 Full-Stack Data Scientists

Shortage of professionals who can build, deploy, and maintain end-to-end ML
pipelines.
Frequently asked Questions
• Will this approach make me industry-ready? Maybe. But not following it will definitely not make you industry-ready!
• Why against learning GenAI? I’m not against it. You should definitely learn GenAI. But without the basics, GenAI won’t
help much. 80% of industry data is still tabular, and most of it doesn’t require GenAI.
• Is statistics and linear algebra absolutely necessary? It depends on your role. You don’t need to be an expert, but
having a basic understanding is essential. Checkout StatQuest with Josh Starmer.
• What’s one mistake I made that others shouldn’t? Not learning MLOps in college!
• GenAI can do coding, conduct data analysis and modelling. Will GenAI take over DS jobs? Not yet.
• What lesson did I learn from seniors at BCG? Understanding what and why to do is more important than how to do it.
Master this, and no technology can replace you.
• AI is changing fast. How do I stay up to date? Everyone has different learning curve. Don’t get overwhelmed by recent
trends. Take it one step at a time, and your efforts will compound.
• What’s your favourite Data Science book and course ?
– Introduction to Statistical Learning by T.Hastie, R.Tibshirani [advance math heavy version, The Elements of Statistical learning]
– Pattern Recognition and Machine Learning by Christopher Bishop
– Course: without any doubt, Andrew Ng Machine learning, Deep learning.
Any other questions??