You are on page 1of 7

5 VALIDATION

AIFL - PROJECT

METR IC

4 P R OJE CT
ME TR ICS

3 MAR KET
R ES EAR CH

2 S OL UTION
AT GL ANCE

1 OPPORTUNIT Y
ANALYSIS
AIFL - PROJECT 1 OPP ORTUNITY
ANALYSIS

P ROBL EM S TATE ME NT L IS T OF PAI N PO INTS

We are a research and advisory firm,


our employees work with alot of Higher Cost

company filings that are in PDF


format/Scanned Images to extract data Accuracy
Resource
Optimization
for research and analysis.

In one such project we process about


240 financial models a year, and each Data extraction
is time
model takes about 10 hours to process consuming Higher work load

of which 7-8 hours is spent on data


extraction
AIFL - PROJECT 2 SOL UTION
AT GLANC E

P RO P OS ED S OL UT IO N
Company filings in PDF format
IN P U T
and Scanned Images

Design a AI tool that will be able to read through


hundrends of PDF document and Scanned Imaged
and give a customised output based on various AI model that can read text data
A I MO D EL
project requirements or client requirements and extract it

PROS CONS
• Reduce Errors • Training AI would be
• Reduce manual effort time consuming on
• Better Resource various unstructured
O U TPU T Data in excel and other formats
Optimization data
3 MAR KE T
R ESE ARC H

DEMAND COMP ET IT ION ANALYS I S


AIFL - PROJECT

Following teams needs help


• Junior Research Analyst
• Senior Analyst
Since this is mostily going to be used Internal, there
• Team using data from PDF
is not competition.
documents
After successful implementation internally, we will
• External Clients that want to
look at creating go to market products.
extract data from documents
AIFL - PROJECT 4 PR OJEC T
METR IC S

Since it is an internal project it will be a cost


centre to the company where the budget will
B U D G ET
be equal to actuals of maintaining resources,
hardware and software.
T ECHNO L O G Y
NLP,CNN,EAST,RCNN,Retinanet,
Tesseract, Hand txt Recognition etc
Extracting incorrect informaion if the model
R IS K is not trainned well with different kinds of
documents
• Product manager
R ES O U R C E
• Data scientist: 1
• Data Engineer - 2
• Deployment expert: 1
T IM EL IN E 3 months
AIFL - PROJECT 4 PR OJEC T
METR IC S

END DATE
START DATE DATE DATE

10th Sept
10 June’21 10th July 10th Aug

100
25% 50% 80%
%

MILESTONE 1 MILESTONE 2 MILESTONE 3 MILESTONE 4

Historic PDF documents NLP, CNN,


EAST,RCNN,Retinanet, Run tests on the models and
of different formats and Deploy the model for testing
Tesseract, Hand txt refine the output
shapes/structures Recognition etc
5 VALIDATION
ME TR IC
AIFL - PROJECT

VA LID ATI O N ME T R IC
All income PDF documents should be extacted at a reduced effort
time
Reduce at least 60% of Manual Effort time
Higher accuracy levels

You might also like