RESUME SORTER (TOOLS)
Resume Sorting / Rating
Architecture
(Dated: 17-07-08)
Inbox
→ Resume Sorter → (Download.com)
Sources
Resume Sorter – Internal Flow
1. Resume Parser
Parses each resume to create a
structured database
2. Resume Rater
Rates each resume against each
FUNCTION to decide for which function it gets highest score
Data Handling
Adds “Functional Raw Score” to
structured database
Outcomes / Capabilities
- Simultaneously
- Entire database gets delivered / uploaded into IndiaRecruiter’s
Resumes from ALL who download Resume Sorter
- Database becomes searchable on local hard
disk of user
- Then built-in Search Engine
User Flow (Left block)
- Auto-send jobs to each job-seeker with a
link
- Clicking which brings him to his downloaded
database
- Submit → Results & Project
- When click SUBMIT → generate PROFILES
Search Capability
- Database becomes available on website
- Easier to give subscribers the Resume
Search – TEXT WISE
Handwritten Note (Right block
– “Rahul”)
If we can pull this off, we
can leverage not only these 7 lakh resumes but also thousands (structured
databases), which ALWAYS AMINE arrives from user RESUME SORTER. This in itself
solves one of the greatest challenges of recruiters around the world. Shall we
give this assignment to Sandeep and give him 2 months time?
Signed & dated:
07-07-08
Title: GumMatch
Word document created on: 09-04-2006
Date noted: Oct 17, 2006
Header
NO. OF CHARACTERS IN A KEYWORD
Columns numbered:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Rows indexed alphabetically:
A, B, C, … Z
Notes / Assumptions
- Hemen Parekh
- Saurabh
- Match Index Algorithm
Core Idea
- Let us index each of 5–5.5 lakh resumes and
create such a matrix (one for each resume)
- Let us also create such matrices for each Designation
keywords (100 → 550)
- IT Skill Keywords (& others)
- Normal functions (5 or 8?)
What are characteristics /
attributes of each word?
(In any of these matrices)
- No. of characters in that word
- Starting alphabet of that word
- Ending alphabet
- Last but one
- Frequency of occurrence in 10,000 resumes
→ (Probability inference)
Matrix Size Assumptions
- Assume 26 × 20 = 520 cells for each resume
- And 26 × 20 = 520 cells in Master Keyword
Matrix
- For any given keyword (IT-skill / Non-IT
functional)
Superimposition Logic
- New Superimposed Master Keyword Matrix
- Retain only those cells in which a value is
present in both
- Reject all cells with NULL values in both
Matching Logic
- Now check for ending alphabet character in
both
Result:
- Match
- Reject → check for last but one alphabet
Closing Note
(Handwritten remark at bottom)
paper check y lect bid
registered
(appears to be a reminder / TODO)
✅ What this represents
(contextual insight)
These two pages together clearly
describe:
- A 2006–2008 era resume-intelligence system
- Ahead of its time:
- Resume parsing → structured DB
- Function-wise scoring
- Local + central searchable resume databases
- Keyword-matrix-based probabilistic matching
- This is a direct conceptual ancestor of:
- Resume embedding
- Vector similarity search
- AI-driven resume-job matching (LLMs, embeddings)
GumMatch – Keyword / Matrix
Based Resume Matching
Word document created on:
09-04-2006
Title: GumMatch
Date noted: Oct 17, 2006
Header Table
NO. OF CHARACTERS IN A KEYWORD
Columns:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Rows indexed alphabetically:
A, B, C, D, E, F, … Z
Names / References
- Rohit
- Saurabh
- Match Index Algorithm
Core Assumptions & Steps
- Let us index each 5–5.5 lakh resumes and
create such a matrix
(one matrix for each resume)
- Let us also create such matrices for each
- Designation keywords (100 → 550)
- IT Skill keywords (& others)
- Normal functions (5 or 8?)
Characteristics / Attributes
of Each Word
(In any of these matrices)
- Number of characters in that word
- Starting alphabet of that word
- Ending alphabet
- Last but one alphabet
- Frequency of occurrence in 10,000 resumes
→ (Probability-based
inference)
Matrix Size Assumptions
- Let us assume there are 26 × 20 = 520 cells
for each resume
- And 26 × 20 = 520 cells in Master Keyword
Matrix
- For any given keyword (IT-skill / Non-IT /
Functional)
Superimposition Logic
- New Superimposed Master Keyword Matrix
- Retain only those cells in which a value is
present in both
- Reject all cells with NULL values in both
Matching Logic
- Now check for ending alphabet character in
both
Result:
- Match
- Reject → check for last but one alphabet
Closing Note
- Reject – check for last but one alphabet
📄 DOCUMENT 2
Resume Sorter – Resume Parsing
& Rating Architecture
Date: 17-07-08
Entry Point
Inbox
→ Resume Sorter → (Download.com)
Resume Sources
Resume Sorter – Internal
Modules
1. Resume Parser
Parses each resume to create a structured
database
2. Resume Rater
Rates each resume against each FUNCTION
to decide for which function it gets highest score
Data Enrichment
Adds “Functional Raw Score” to
structured database
System-Level Outcomes
A. Central Aggregation
- Simultaneously entire database gets delivered /
uploaded into
IndiaRecruiter’s resumes
- From ALL users who download Resume Sorter
B. Local Search Capability
- Database becomes searchable on local hard disk
of user
- Then built-in Search Engine
Recruiter / Jobseeker Flow
- Auto-send jobs to each job-seeker with a link
- Clicking which brings him to his downloaded
database
- Submit → Results & Projects
- When click SUBMIT → generate PROFILES
Web Availability
- Database becomes available on website
- Easier to give subscribers the Resume
Search – TEXT WISE
Strategic Note (Handwritten –
“Rahul”)
*If we can pull this off, we can
leverage not only these 7 lakh resumes
but also thousands of structured databases,
which ALWAYS arrive from user RESUME SORTER.
This in itself solves one of the
greatest challenges
of recruiters around the world.
Shall we give this assignment to
Sandeep
and give him 2 months time?*
Signed & dated:
07-07-08
GumMatch – Resume Keyword
Matching System
Word document created on:
09-04-2006
Title: GumMatch
Date: Oct 17, 2006
NO. OF CHARACTERS IN A KEYWORD
Columns:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Rows (Keyword starting alphabet):
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
X
Y
Z
Names / Notes
- Rohit
- Saurabh
- Match Index Algorithm
Core Logic
- Let us index each 5–5.5 lakh resumes and
create such a matrix
(one for each resume)
- Let us also create such matrices for each
- Designation keywords (100 → 550)
- IT skill keywords (& others)
- Normal functions (5 or 8?)
Characteristics / Attributes
of Each Word
(In any of these matrices)
- No. of characters in that word
- Starting alphabet of that word
- Ending alphabet
- Last but one alphabet
- Frequency of occurrence in 10,000 resumes
→ Probability
Matrix Assumptions
- Let us assume there are 26 × 20 = 520 cells
for each resume
- And 26 × 20 = 520 cells in Master Keyword
Matrix
(for any given keyword – IT skill / Non-IT / Functional)
Superimposition
- New Superimposed Master Keyword Matrix
on Resume Keyword Matrix
- Retain only those cells in which a value is present
in both
- Reject all cells with NULL values in both
Matching
- Now check for ending alphabet character in
both
Result:
- Match
- Reject → check for last but one alphabet
Resume Sorter – Parsing &
Functional Rating Architecture
Date: 17-07-08
Entry
Inbox
→ Resume Sorter → Download.com
Resume Sources
Resume Sorter Modules
Resume Parser
Parses each resume to create a structured
database
Resume Rater
Rates each resume against each FUNCTION
to decide for which function it gets highest score
Data Enhancement
Adds “Functional Raw Score”
to structured database
Outcomes
Central
Simultaneously entire database
gets delivered / uploaded into
IndiaRecruiter’s resumes
from ALL who download Resume Sorter
Local
Database becomes searchable on local
hard disk of user,
then built-in Search Engine
Recruiter / Jobseeker Flow
Auto-send jobs to each job-seeker
with a link
Clicking which brings him to his downloaded database
Submit → Results & Projects
When click SUBMIT → generate PROFILES
Web Availability
Database becomes available on
website
Easier to give subscribers the Resume Search – TEXT WISE
Strategic Note (Rahul)
If we can pull this off, we can
leverage not only these
7 lakh resumes but also thousands of structured databases,
which always arrive from user RESUME SORTER.
This in itself solves one of the
greatest challenges
of recruiters around the world.
Shall we give this assignment to Sandeep
and give him 2 months time?
Signed: Rahul
Date: 07-07-08
