Tickets

Loading…

Tushar Garimalla

@tushargarimalla

AI Credit Manager that makes loan decisions

Submitted Oct 11, 2026

{Submission type - choose one:

  • Talk/session proposal - 30-40 mins
  • Showcase - new ideas - 20 mins
  • Poster - alleyway
  • Problem/reverse CfP - potential Birds of Feather
  • Problem exchange facilitator/helper - volunteer/find collaborators}

If talk/session proposal, showcase, or poster
Lending predates money! Formal loan evaluation (banks / NBFCs) was always done by a loan officer / credit manager who manually reviewed every application and made loan decisions. As retail lending expanded and loan sizes reduced, the cost of the credit manager became prohibitive to each loan. Data science stepped in to ensure that loan evaluation cost (opex) was always low. However, all models inherently have error and drawing any cutoff always comes with trade-offs. Retail risk manager expertise in the last 20-30 years has focused on managing model risk through various business and macro cycles.

AI presents an opportunity to rethink this approach. Hypothetically if the chief risk officer of a bank evaluated every loan manually, the banks “highest level of risk expertise” is getting used for every application. So, instead of or in conjunction with data science, the CRO could train an AI credit manager (like how the credit org trains freshers / junior hires over a period), assign decision authority and review performance of AI credit managers. Given certain unique abilities of AI over human and data science model capabilities, the overall risk org is likely to make better decisions and even open up certain segments that are currently considered un-lendable.

Key Takeaways

  1. AI presents an opportunity to rethink retail loan evaluation
  2. Implementation, review, compliance and governance will have to be re-built with a first principles approach
  3. Business gains are likely but needs capital to test at scale

Anybody in lending or data science will benefit from this session.

https://www.linkedin.com/in/tushargarimalla/
IITM 2005 B.Tech (Elec)
Credit risk and lending experience in US and India
Managed Best Buy credit card portfolio at HSBC USA and Capital One USA
Built Amazon Pay Later at Capital Float India
Co-founded Aspire NBFC in India focuses on working capital for shopkeepers

Link to the deck: https://drive.google.com/file/d/1Wnh5vtEhHUH-S-r1dLHqTA11M_su7mxy/view

What we don’t know yet is a way to quantify the benefit of AI over “traditional” data science w.r.t. business gains

Match making - anyone who is interested in opening up underserved segments using AI

Domain tags - fintech, financial inclusion, lending

If problem/reverse CfP
{Describe the problem in one short card: what is broken, blocked, unclear, or worth exploring?}

{What kind of help would be useful? Choose any that apply:

  • idea
  • experience
  • technique
  • dataset
  • collaborator
  • critique
  • domain context
  • implementation help}

{Topic/domain tags - choose 2-3 broad tags that describe your problem space.
Examples: data for good, causal inference, synthetic data, computer vision, fraud, ecology, healthcare, governance, evaluation, safety, education, public sector, climate, finance.}

{Your role as problem owner: by submitting this, you are the default owner of this problem card. Are you willing to be contacted before the event for matching or clarification?}

{Add name, affiliation, and short bio. Make it intriguing!}

If problem exchange facilitator/helper

{How do you want to help? Choose any that apply:

  • rewrite vague problem statements
  • merge duplicate or overlapping problems
  • suggest matches between people/problems
  • facilitate a problem exchange session
  • introduce someone who may be useful
  • provide subject-matter critique}

{What can you help with? Choose any that apply:

  • ideas
  • experience
  • technique
  • dataset
  • collaborator intro
  • critique
  • facilitation}

{Topic/domain tags - choose 2-3 broad tags that describe where you can be useful.
Examples: data for good, causal inference, synthetic data, computer vision, fraud, ecology, healthcare, governance, evaluation, safety, education, public sector, climate, finance.}

{Availability: are you available before the event for light-touch matching/facilitation?}

{Add name, affiliation, and short bio. Make it intriguing!}

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

One year entry to The Fifth Elephant's programmes and activities

Hosted by

Jumpstart better data engineering and AI futures