Leonid Nekhymchuk: LLMs and No-code, how to simplify data operations with AI (UA)
Sep. 9, 2023•0 likes•23 views
Report
Business
Leonid Nekhymchuk: LLMs and No-code, how to simplify data operations with AI (UA)
AI & BigData Online Day 2023 Autumn
Website – www.aiconf.com.ua
Youtube – https://www.youtube.com/startuplviv
FB – https://www.facebook.com/aiconf
2. 2
LEONID NEKHYMCHUK
CEO & Co-Founder @ Datuum.ai
Software developer turned data expert
passionate about building data products with AI.
3. 3
Is LLM the biggest technology breakthrough since the Internet?
“Research estimates that generative AI could add the equivalent of $2.6 trillion.
to $4.4 trillion annually across the 63 use cases we analyzed”.— McKinsey.
Threads
ChatGPT
Instagram
Spotify
Dropbox
Facebook
Foursquar
e
Twier
Airbnb
Kickstarter
Netflix
0 500 1,000 1,500
Online
Service
Time taken to reach 1 million users (Days)
Time Taken To Reach 1 Million Users
30.0%
20.0%
10.0%
0%
-10.0%
-20.0%
Monthly Visits Desktop & Mobile Web Worldwide YOY
M
ay
2021
J
u
l
y
2
0
2
1
S
e
p
2
0
2
1
N
o
v
2
0
2
1
J
a
n
2
0
2
2
M
a
r
2
0
2
2
M
a
y
2
0
2
2
J
u
l
y
2
0
2
2
S
e
p
2
0
2
2
N
o
v
2
0
2
2
J
a
n
2
0
2
3
M
a
r
2
0
2
3
github.com
stackoverflow.com
4. 4
Who needs to get ready for the impact
Using generative AI in just a few functions could drive most of the technology’s impact across potential.
corporate use cases..
Impact, $ billion
Impact as a percentage of functional spend, %
Represent ~75% of total annual impact of generative AI
5. 5
Data world — how LLM can help
Task
Data
Prompt LLM Code Validation
Metadata
1 Understanding semantics of your data
2 Map your source to your destination
3 Generate ETLcode
7. LLM and no-code
7
No-code interface - harnessing the power of LLM for non-technical users
Datuum — data transformation examples.
“Why LLM doesn’t replace developers …yet.
8. LLM in the corporate world
8
Everyone
wants their
own ChatGPT Private
data
Security
considerations
9. 9
LLM and private data - Retrieval Augmented Generation
Prompting
PDF
PDF
LLM (GPT)
Create embeddings
from pdf files
Save embeddings
in a vector database
Get chunks of text based
on vector similarity search
Text chunks, memory,
context and optimized prompt
Answer based on PDF sources
and conversation history
Retrieve and store
conversation history
Answer
Prompt: What is
Photosynthesis?
Business
user
LangChain
Conversation
Chain
Vector
Database
Memory/History
Storage
Typical use-cases: search, summarization, Q&A
Everyone wants an LLM that understands its data
Typical architecture:
Preparation