TechX

"We are here to level up your Tech Game"

 TechX
presents
+ AI/ML Sig

Building Value Streams

through “Data”
Types of Data :
1. Structured Data 2. Unstructured & Semi-Structured Data
 Names  Images (human- and machine-generated)
 Dates  Video files
 Phone numbers  Audio files
 Currency or prices  Social-media posts
 Heights or weights  Product reviews
 Latitude and longitude  Messages sent by SMS or through online services.
Types of Data :
1. Qualitative Data 2. Quantitative data
 Nominal Data  Discrete Data
 Ordinal Data  Continuous Data
 Database
1. Structured Data 2. Unstructured & Semi-Structured Data
 SQL  No SQL
 My SQL  Object Storage (Cloud)
TECHNICAL ARCHITECTURE

Google
Analytics

Google
ML

Data Replication
(Backup)
 PROCESS FLOW
Data Data ETL Data Science Data Live User
Web-App
Collection Cleaning process ML Analytics Monitoring Feedback

Real-time Monitoring is Web-App is the central USER FEEDBACK

We will collect all forms Now, we would need to ETL (Extraction, The real Game begins here: essential in the initial point of contact for user must be carefully
of data both Structured clean the data we Transform & Load) is an Data now needs to be analyzed and phases of the product interaction. Getting this monitored & this step
and Un-structured data, ingested from various important step in parallelly fed into different models for development, as a on point is very vital for is more like food to
such as tabular data streams. This is a making sure our final testing and prediction. Post this we need human needs to make the businesses and the whole system.
relational) data, IOT MAJOR step as Good data is ready to be to create GAN/NN models for GEN-AI. sure qualitative inputs future growth. Keeping Qualitative feedback
sensor data (time- data drives Good mashed up against our are being fed into it Simple yet Efficient is empowers the whole
series), and blob results. various data sources &
Products needed: system for the model the game changer. process and
data(images, audio, INTEGRATION between• Data Science platforms accuracy & training.
improves the
tweets) etc. Products needed: • Jupyter / Anaconda
these sources should Products needed: efficiency of feature.
• Lot of Opensource occur organically • Powerful GPU backed platforms (Cloud) Products needed: • Oracle Apex
Products needed: technologies like without any errors.• Data Analytics platforms • Cloud Monitoring • Low code platforms Products needed:
• Oracle RDBMS (SQL SQL, NoSQL, Python • Selenium
• Chatbot
- Structured data) • Open Refine by Products needed: Skills needed: • Cypress Skills needed: • CSS
• NoSQL/Hadoop (Un- Google • Data Warehouse • Python Skills needed: • Java • WebApp
structured/semi- systems & • R • Java • Python Skills needed:
structured data) Skills needed: Integration services • SQL • Python • C# • Java
• Cloud Object • SQL/PLSQL from Cloud Vendors • Tableau / OAC • C# • SQL • Python
Storage • NoSQL • Enterprise S/W ETL • Business Understanding • Feature • CSS • C#
Skills needed: • Python • Hadoop Understanding
• SQL/PLSQL Skills needed:
• NoSQL • Java PS: More on this specific part on further
• Python slides.
• Parameterization
Basic GenAI Architecture

Source: https://www.accenture.com/us-en/blogs/cloud-computing/building-generative-ai-we-can-trust
Training the model

Source: https://youtu.be/G2fqAlgmoPo?si=JFl78cg2LL9DwuDO
 Prediction/ Application
The various devices linked to
delivering the enhanced customer
experience e.g. AC, seat adjustment
actuators, music systems, etc.

Testing Dataset
The Test Data
Usually, 30% data is kept for
training the algorithm

Training Dataset Algorithm Validation Model

Data
The Training Data
Usually, 70% data is kept for
training the algorithm

Input Data
The Data The various devices linked to
The core data received form the delivering the enhanced
images of drivers, Input form customer experience e.g. AC, seat
ECU, music systems is used to adjustment actuators, music
train and tune a foundation systems, etc.
model
Thank you…
Boost - Best way to kickstart the GenAI game

A PRE-TRAINED GENERIC PURPOSE

BASED ON THE USE-CASE, THE
LLM MODEL WILL BE BOUGHT MORE AND MORE REAL TIME DATA,
MODEL WILL BE TRAINED WITH A
FROM LARGE VENDORS SUCH AS NEWS, TRENDS, ETC. WILL BE
SMALLER BUT SPECIFIC DATASET
GOOGLE WHICH WILL EQUIP OUR FEEDED TO INCREASE THE
SUCH AS FINANCE AND STOCK
GENAI TO DO BASIC HUMAN-LIKE ACCURACY AND PERFORMANCE
MARKET DATA
CONVERSATION

Source: https://youtu.be/G2fqAlgmoPo?si=JFl78cg2LL9DwuDO
Why pre-trained APIs?

Source: https://youtu.be/G2fqAlgmoPo?si=JFl78cg2LL9DwuDO
 Application in the Automotive
Industry

• Generate new • Generate • Personalize • Optimised

Safety

Customer Experience
Design

Supply Chain
design concepts realistic infotainment Production and
• Create virtual simulations of systems inventory
prototypes accidents, ---> • Create plans-- Cost
• Manufacturing train customized minimization,
process autonomous driving modes waste reduction
optimization vehicles • Voice • Analyse sensor
• Identify recognition data on
potential safety capabilities of production
hazards in in-car virtual line--- detect
existing assistants anomalies and
vehicles---> prevent it from
Reduce reaching
Accidents customers