[DSC DACH 25] Ivan Vukasinovic $20M Datapipeline Built by AI in 48 hours.pdf

What if an AI could design, optimize, and deploy a multi-million-dollar data pipeline — in just 48 hours? In this session, Ivan Vukasinović, CTO at MILLIONS.co, shares how his team leveraged cutting-edge AI tools to build a data infrastructure powering $25 million in business operations. From automating architecture decisions and resource allocation to optimizing performance and cost, this talk reveals how AI can accelerate development cycles and redefine what’s possible in data engineering. Aimed at intermediate-level professionals, the session offers a behind-the-scenes look at how AI-driven automation can deliver massive business value in record time.

1. $20M DATA PIPELINE
BUILT WITH AI IN 48 HOURS
Presented by
Ivan Vukasinovic
CTO @ MILLIONS CO.
https://millions.co [email protected] Toronto, ON Canada

2. Social Campaign Efficiency
BUSINESS GOALS
Talent Post Performance Talent Audience Info

3. Zero to a Fully-Featured, Serverless DaaS Platform 48 hours
ENGINEERING CHALLENGE
Challenged Everyone I ever worked for the past 15 years....
Impossible, you need days of data analysis, researching data sources
Its going to take you days of devops, if not weeks....
The architecture design is very complex for DaaS Platforms, AI can’t
help you.

4. KISS – Keep It Simple Sunshine
Yes, current modern day AI programing tools cannot do
complex bussiness logic and complex solution architecture to
understand and generate production ready code for these
particular usecases.
STRATEGY
AI EMPOWERED
Serverless Micro-service Decoupled Architeture
Breaking the data pipeline into smaller pieces that are
decoupled and using the serverless framework eliminates the
need for complex Devops work. Serverless is simply enough yet
can run any codebase, python is the best match due to
simplicity and great support in data processing frameworks.
Divide and Conquer – Cursor, Gemini and ChatGPT Unite
Leveraging multiple AI tools, Gemini and ChatGPT to evaluate
data providers and run basic data analysis as well as provide
sample curl calls to get the data for Cursor. Cursor having the
architecture diagram as well as a sample calls as well as usecase
exampled alongside rulles created for code generation acts way
above any intermediate developer if not a senior one

5. GEMINI – FINDING THE DATA
Finding possible data sources that are
alternative to current one to fit the bussiness
needs and financial requirements. Checking
data in DynamoDB and ElasticSearch for
quality and proper processing.
DIVISION OF LABOUR
CURSOR – BUILDING THE PIPELINE
Building the actual pipeline piece by piece, one
lambda at a time, then connecting to external
data sources with API Keys safely in AWS
SecretsManager and storing and data in
DynamoDb and Elastic Search. Python.
CHATGPT – ARCH BRAINSTORM
Analysing the architectural needs for the
solution, which tools to use understanding cost
analysis and breakdown for the SLAs.
HUMAN – FINAL ASSEMBLY
Reviewing results from all 3 AI tools, fine tuning
it and putting pieces together, taking diagrams
from ChatGPT and data from Gemini and
feeding it into Cursor. Running Cursor code on
AWS to deploy and run the DP.

6. THE PLATFORM – SERVERLESS DATA PIPELINE PROCESSING INSTAGRAM DATA IN 48 HOURS

7. Our success was driven by specific, contextual, and
chained prompts.
We acted as senior engineerings guiding the AI like a
junior engineer through complex tasks correcting it when
it makes a mistake.
First Prompt: "Write a Python Lambda handler that
triggers from S3, calls the Apollo API, classifies the result
with OpenAI, and writes the enriched data to
DynamoDB..."
PROMPTS TO CODE

8. It would require a data engineering team of
multiple data engineers, a devops, data
analyst, etc...
Traditional Timeline – 3 Months
TRADITIONAL VS AI POWERED
We use the latest tools and technology to
stay ahead, ensuring efficient and effective
services.
Traditional Cost – more than 100K USD
Leveraging knowledge from 15 years of
building data pipelines and cutting adge AI
Tools allowed one domain expert to build
this in a fraction of the time.
AI Powered Approach – 48 hours
If we assume the exter level time cost and
the AI tools rough cost is fraction of
traditional cost.
AI Powered Cost – less than 5K USD

9. 100s of thousands dollars saved
10% margin increased in our B2B SaaS Product
10+ new bussiness features supported
THE BUSSINESS IMPACT We went from initial concept to a deployable,
production-grade architecture in a single weekend.
This resulted in a 6+ month acceleration in time-to-
market and over 90% greater capital efficiency.
HTTPS://CAMPAIGN.MILLIONS.CO

10. One Man Unicorn
The 1-Person Unicorn:
With these tools, it is now
conceivable for a single
founder to build a highly
complex and valuable
software company.
LESSON LEARNED – EVOLVE OR EXTINCT
AI Power Human Made
This entire serverless
platform was architected
and coded in 48 hours.
The future of engineering
is a powerful collaboration
between human vision
and AI execution.
Devops Friction
40% of our time was
spent deploying to AWS
as the AI Tools still don’t
integrate well with cloud
platforms but when they
do it will he a
gamechanger.
Relying on traditional
development today is like
bringing a slingshot to a
starship battle. AI isn't just
an upgrade; it's an entirely
new class of technology
that has fundamentally
changed the rules
Evolve or Extinct

11. THANK YOU Presented by
Ivan Vukasinovic
CTO @ MILLIONS CO.
https://millions.co [email protected] Toronto, ON Canadafor your time and attention