Friday 9 Oct 2026, 4 to 8 pm. Two parts: a session on banks, RBI and NBFCs, then a demo of FinStack's platform and the hackathon brief.
Written in my own words from my notes of the recording. Items marked [verify] were said in class in a simplified way and need checking before I repeat them.
Part 1: how banks, RBI and NBFCs work
The big idea. A bank trades money: it buys money (deposits, capital) and sells it (loans) at a higher rate. RBI sets rules so a bank never runs out of cash.
Assets and liabilities flip for a bank. My loan is my debt, but it is the bank's asset because it earns interest. My fixed deposit is my asset and the bank's liability, because it pays interest and I may withdraw it.
Three safety buffers
| Buffer | What it is | Earns the bank |
|---|---|---|
| CRR | Cash kept with RBI, cash only | Nothing |
| SLR | Government bonds, gold and other very safe assets | Stable, low |
| HQLA | Liquid assets to cover the next 30 days of expected outflow | Some |
About a quarter to 30% of deposits sit in these buffers; the rest can be lent. The reason is to stop a bank run, where panic makes everyone withdraw at once even if the bank is fine (Silicon Valley Bank and Yes Bank were the examples).
Every loan creates a deposit. When a bank lends, the money lands in an account at the same bank. Nothing physically moves; one balance becomes two entries. That is why banks build ecosystems (apps, salary accounts, no-cost EMIs): to keep the money inside.
LAF, repo and reverse repo. Under the liquidity adjustment facility a bank swaps cash and bonds with RBI. The repo rate is what a bank pays to borrow cash from RBI; the reverse repo rate is what RBI pays a bank that parks cash. Lower repo means easier lending; higher repo cools inflation.
NBFCs. They lend but, in the deposit-taking case, follow similar ratios and pay a little more on deposits with less insurance cover. A cheaper source of money lets one NBFC serve good borrowers; a dearer one must lend to riskier segments at higher rates. When two lenders chase the same segment they compete on how easy the onboarding journey is.
What I took away: the lender's cost of money decides which customers it can serve, and when costs are equal the journey itself is the product.
Part 2: what a no-code lending platform is
What FinStack is. Software sold to lenders (banks and NBFCs). It does not lend and does not tell the lender what rules to set. The lender's own customer is the borrower; FinStack's customer is the lender.
Origination vs servicing. Origination (sourcing, KYC, underwriting, decision) is where lenders can differ. Servicing (EMIs, reporting, collections) is heavily regulated, so there is little room to differ. The hackathon is about origination.
Why lenders buy instead of build
- Integrations: Aadhaar, PAN, bank statements and bureaus need Indian vendors, and generic no-code tools do not have them.
- Compliance: an audit trail of every action, masked data, consent records.
- Speed: a risk team can change a rule in seconds, not months.
- Data stays with the lender, deployed in the lender's own cloud account.
The building blocks
| Block | What it does |
|---|---|
| API hub | Ready verification APIs, each an input and an output; several vendors per check, with fallback |
| Form builder | Forms are the building block of a journey: API forms (PAN, bureau) and custom forms |
| Journey builder | Orders the forms into steps; sets what happens on pass or fail |
| Rules | Verification rules (for example a fuzzy name match at 80%), visibility rules, logic variables |
| Internal workflow | Back-office screens for review and approval |
| Audit trail | Every entry, API response and decision logged |
Underwriting in one line. Decide whether to lend, how much and at what rate. Knock-out rules (for example a score under 650 or a name mismatch) reject at once.
Bureau and Account Aggregator. The platform cannot make hard credit pulls or Account Aggregator calls on its own; the licensed lender supplies the credentials. PAN and Aadhaar checks work without that.
Privacy law (DPDP Act). Take consent and say why the data is collected, encrypt and mask personal data, delete it when the consent period ends.
The hackathon brief
- Build any fintech product on the form and journey builders.
- KYC is mandatory; put a consent tick in the first form; onboard an individual, not a business; use soft-pull bureau calls only.
- A complete simple product beats a half-built big one.
- Teams of two, one technical and one not.
What I took away
- Banks and NBFCs differ by cost of money and by chosen customer segment. The segment decides the rules.
- In a lending product the order of checks is a design choice. Cheap checks go first, costly ones later.
- A platform like this turns "build an app" into "configure forms, order them and add rules". The thinking moves to the customer, the rules and the data.
Where I used it in DealerFinance
- Consent tick first, then the checks (Step 1).
- Lender fit rules per lender, as simple thresholds (the knock-out idea).
- A journey of forms in an order, with a rejection sending the case back (journey builder).
Open or to verify
- Deposit insurance cover as said in class does not match what I know. The current cover is ₹5 lakh per depositor per bank [verify].
- The CRR range said in class was 3 to 5% [verify the current level].