Hobbiate, 136 A EC Road, Dehra Dun (2026)

30/03/2026

At a time when the world feels uncertain, one thing remains constant: Lending is the backbone of economic resilience.

In South East Asia, lending isn't just a financial activity—it’s how local economies stay alive. But the "growth at all costs" era is over.

The next phase of winner-takes-all in lending won't belong to the reckless. It will belong to those who can be efficient and conservative simultaneously.

The challenge?

Capital is still in high demand.

But mistakes are getting more expensive.

Every credit decision carries more weight.

This isn't a phase for volume chasing. The strongest lenders will be those who can:

Move fast without becoming loose.

Stay prudent without becoming slow.

Scale volume without inflating operating costs.

In today's market, conservative lending isn't the opposite of growth—it’s the foundation of sustainable growth.

18/03/2026

Necessity doesn’t just create innovation. Sometimes it forces us to abandon bad habits.
Recently in Thailand, due to an energy crisis, offices reduced air-conditioning to save electricity. Something interesting happened: Journalists and officials stopped wearing suits.
It wasn't a fashion decision. It was a survival decision. Without aggressive AC, wearing a wool suit in a tropical country becomes a form of physical punishment.
It reminded me of something I have often observed in Mumbai. If you step outside for five minutes in the afternoon, you are drenched in sweat. The climate is hot and humid. For centuries, people here wore clothing that worked with the geography—loose cottons, lungis, and dhotis.
Then, colonial corporate culture arrived with wool suits designed for London winters. Somehow, the ritual stayed.
An entire ecosystem was built to sustain this irrationality: The Suit → requires strong AC → The AC justifies the Suit.
Remove the AC, and the logic collapses instantly.
I’ve spent time in Dehradun and Bangalore, where most homes don't have (or need) air conditioning. We are perfectly fine at 32°C or 35°C with a simple ceiling fan. Yet, the moment we enter a corporate office, we are expected to live in an "Arctic" 20°C.
Why? Not for the human, but for the jacket.
There is a profound sustainability angle here. When we stop cooling buildings to "suit-temperatures" and adopt climate-appropriate habits, we save enormous amounts of energy.
This leads to a beautiful trade-off:
We can keep our factories running.
We can keep our homes lit.
We can celebrate Diwali and Holi with all their tradition.
And still leave breathing room for the planet.
Sometimes sustainability doesn't require a billion-dollar "Green Tech" innovation. Sometimes it just requires letting go of an irrational habit.
Personally, I don’t oblige the inappropriate demands to "appropriately" dress for this "suit-boot" industry. I prefer to dress for the world I actually live in, not the one colonial history told me to inhabit.
Necessity is the mother of invention, but it is also the mother of leaving a faulty culture behind.

14/03/2026

# # The "Shiny Tool" Trap: Why Your Automation is Still Manual

On the first week of my software engineering career, my team was given ropes, nails, and stools for a team-building exercise. The goal? Deploy an object at the top of a tree.

Every team that tried to use the "provided tools" failed. They were so focused on how to use the rope that they forgot to look at the tree. Finally, one team ignored the tools entirely, innovated a new path, and succeeded.
Most of the time, having a tool causes us to use it even when we don’t need it. We take the elevator for two floors and the car for two blocks, just because the tool exists.

I saw a high-stakes version of this in Fintech recently.

A friend was boasting about a "foolproof" tool for bank statement analysis. It caught tampering, parsed data instantly, and was supposedly the future of their credit department.

Then we found the trap: 80% of their incoming documents weren't digital PDFs. They were messy, tilted, low-resolution scanned statements.
The result? The "shiny tool" couldn't read them. So the company had two choices, both of them bad:
- Reject 80% of their potential customers.
- Build a massive "manual processing" department to do the work the tool couldn't.

They bought a high-speed elevator for a building that only had two floors.
At Hobbiate, we built FinLens with the "tree" in mind, not just the "rope." We knew that real-world credit doesn't happen in perfect digital files. It happens in the 80%—the scans, the photos, the messy reality.

If your "automation" still requires a room full of people to handle the "exceptions," you don't have a tool. You have a bottleneck with a fancy name.

Don’t build your process around the tool. Build the tool around the messy truth of the problem.

hashtag hashtag hashtag hashtag hashtag hashtag hashtag hashtag

23/02/2026

The Dairy, the Guard, and the 1,200-Page File
A friend who runs a dairy once told me a story about missing milk. 🥛
He noticed the numbers weren’t adding up, so he hired a security guard.
The next day, milk still went missing.
So he hired a second guard to watch the first.
When that didn’t work, he hired a security manager to oversee the guards.
The payroll went up.
The milk still disappeared.
Then he finally did the one thing that worked: he installed a CCTV camera.
Suddenly, he didn’t need more eyes.
He needed one source of truth.
In many countries, lending is still stuck in the “security guard” phase.
When defaults rise, the instinct is to hire more analysts.
When that doesn’t help, firms add QC layers and more managers to “eyeball” the work.
Pretty soon, the cost of underwriting is eating into interest income—and the risk is still there.
I saw this recently with a 1,200-page bank statement set.
A human trying to “eyeball” 1,200 pages is just a guard standing at the gate.
If they have to skim to meet a quota, they aren’t really seeing risk—they’re just watching the clock.
FinLens is the CCTV.
It processed those 1,200 pages without getting tired.
It didn’t “trust” the person carrying the milk.
It recorded every transaction, verified every balance, and flagged anomalies with cold, deterministic precision.
You don’t solve leakage by adding more eyeballs.
You solve it by installing a system that doesn’t need to blink.
If your underwriting costs are rising but your defaults aren’t falling, you probably don’t need more guards.
You need a camera.
hashtag hashtag hashtag hashtag hashtag hashtag hashtag hashtag

09/02/2026

I used to think bank statement analysis wasn’t relevant to the US market.
Turns out, as a function, Bank Statement Analysis keeps throwing surprises- even to someone with 10+ years of deep experience in this space.

We recently noticed how some US statements quietly encode tier / product identifiers in formats, fees, and transaction patterns. These little details change how data should be interpreted—and if you ignore them, your “smart” system will confidently get things wrong.

Real fintech AI isn’t about fancy labels. It’s about surviving the messy, undocumented reality of financial data.

23/01/2026

Over the last year, something interesting has happened.
Some of our enterprise customers signed up without a demo.
No trial. No pilot. Just a few conversations. And then- production.
At first, this felt unusual.
But then it made sense.
When a team has already lived the pain- messy bank statements, scanned documents, passbooks, operational risk- a demo doesn’t reduce uncertainty.
It sometimes adds it.
What they really want to know is:
Has this worked for someone like us?
Where does automation stop?
What happens when the system is unsure?
For risk-heavy, ops-heavy problems, trust travels peer to peer, not screen to screen.
Demos sell excitement.
Production systems sell sleep.
We’ve started respecting that distinction a lot more.
Curious- have you seen similar patterns in enterprise or infra products?

23/12/2025

Merry Christmas 🎄 & Happy New Year 🌱

Warm wishes to everyone.
May this festive season bring peace, good health, and renewed hope.

And may the year ahead help humanity become a better version of itself.

06/12/2025

This week, no AI talk.

Winter + Hot Toddy = honesty mode.

So here’s a story I’ve never shared- the secret sauce behind BharosaAI.

A few years ago, I sat down to write a complex module for FinLens.
I opened the file…
looked at the logic…
thought of all the permutations, edge cases, and data variations…
And I honestly told myself, “Main to buddhu hoon — I can’t mentally test such complicated code. If I don’t have testcases, nothing makes sense to me.”
So instead of writing code,
I spent five full days just thinking and writing testcases.
No code.
No shortcuts.
No ego.
Just clarity.

And something unexpected happened.
Those testcases became the foundation of our engineering culture.
Suddenly:
- Code became simpler
- Boundaries became clear
- Modules stopped interfering with each other
- Developers couldn’t accidentally break flows
- The system started policing itself
- If someone skipped a standard, I didn’t have to say anything.
- Tests screamed.
- Flows broke.
- And the engineer realigned.
- Automatically.

Recently a developer said:
“Sir, aapka play-button concept sahi hai.
Click karo, sab sahi ho jaata hai.
Ab kuch bigaadne ka chance hi nahi.”. (Sir, you play button concept is great, just click it and everything happens, no chance of going back).

That’s when I realised:
- This isn’t just engineering.
- This is Bharosa.
- You cannot build BharosaAI
if your own system doesn’t have Bharosa in itself.

And since I’m already in honesty mode…
If you adopt this discipline-
test-first, clean boundaries, predictable workflows-
- You will ship faster than us.
- You will build more stable systems.
- You will probably beat us.

Because this is the real secret sauce.
AI baad mein aata hai.
Pehle Bharosa banta hai.
(Blame the Hot Toddy for this post.)

Happy building.
Happy testing.
Happy winter-honesty-mode.
🔥🥃🚀

hashtag hashtag hashtag hashtag hashtag hashtag hashtag

17/11/2025

⚙️ “Not now.” Until… Now.

When LLMs first arrived, people said —
“Let’s do something with this!”
I said, “Interesting. Not now.”

When agentic workflows appeared, they said again —
“Let’s build something fast!”
I said, “Interesting. Not now.”

While we were already delivering agents to clients, some said,
“Why not make one for ourselves?”
I smiled — “Not now.”

Recently, a prospect said,
“Show me all your agents.”
And I thought, “Oh no, not another hello-world agent.”
Still… Not now.

Then came the right moment.
FinLens was stable in production.
My team asked, “What next?”
I said — “BharosaAI. The future platform.”

For months, we built the foundation quietly.
Until one day at lunch, an idea struck.
I turned to my teammate and said,
“Let’s stop all FinLens work for now and go full steam on BharosaAI.”
She asked, “Next week?”
I said — “No, NOW.”

💭 Reflection
We don’t pick problems because they’re trending.
We pick them when we know we’re about to change how things fundamentally work —
and when we’re confident that what we build cannot be easily replicated.
For us, “Not now” isn’t hesitation.
It’s discipline.
Because when the moment truly arrives —
we don’t say “Let’s start.”
We say — “Now.”

10/11/2025

🌾 “Not now.”

Around 2005, I wanted to build a company where people could make a list of groceries and get them delivered to their homes.
Then I thought, “Not now.”

Years later, BigBasket made it happen.

Then I imagined sensors on every kitchen box, reordering things automatically when they ran out.
Again, “Not now.”

I once dreamt of an AI that could take anyone’s voice and sing an entire song in it.
Looked around, said — “Not now.”

Then came Suno AI.

I wanted to blend two rāgas, use AI to create new melodies.
But I was building FinLens and other BFSI systems — so again, “Not now.”

And now, I’m thinking of AI-generated movies.

🎬 The way things are going,
the biggest movie producers might not be Hollywood studios at all —
they might be TCS, Infosys, or Accenture.

Because the future of creativity is not just art — it’s compute.
When content becomes code, those who manage compute will manage culture.

There’ll be a few years of chaos — copyright battles, licensing confusion —
but it’ll settle.

🎶 For songs, the money will still be in concerts.
Humans will still fill halls.

🎭 For movies, “Avatar-like” experiences will be normal.
Actors will still matter — not just for their faces, but for their presence.

💭 And here’s the truth:
We still have these opportunities.
The world hasn’t finished evolving — it’s looping back with better tools.

Every time I said “Not now,”
it wasn’t hesitation.
It was preparation.

✨ If you’ve ever paused an idea because the world wasn’t ready —
remember, maybe you were just early.

03/11/2025

Everyone wants to be “AI-enabled.”
Because the fear is real — *What if someone new uses AI to replace us?*

But most AI projects don’t fail because the technology doesn’t work.
They fail because the **intent** wasn’t clear.

Before starting any “AI initiative,” pause and ask your team these eight questions 👇

1️⃣ What problem are we solving that isn’t already solved?
2️⃣ Why is AI the right tool for this?
3️⃣ Who owns the outcome, not just the model?
4️⃣ Do we have the right data — and permission to use it?
5️⃣ Is success measurable?
6️⃣ Who’s being replaced — or augmented?
7️⃣ Are we fixing the workflow, or just adding a layer?
8️⃣ Can we sustain it beyond the demo?

At **Hobbiate**, we faced these same questions while building *FinLens* — and those lessons shaped how we approached AI itself.

That mindset led to *BharosaAI* — not as a silver bullet, but as a framework that brings **discipline, auditability, and orchestration** to how AI runs inside an enterprise.

Every single line of code in BharosaAI justifies its **direct contribution to FinLens**,
unless a services customer asks for a feature — in which case BharosaAI extends itself to a *real, proven use case.*

Because AI doesn’t need more enthusiasm.
It needs **trust, structure, and intent.**

🧠 Think before you automate.
🎯 Simplify before you optimize.
🚀 And only then, scale with AI.

AI done with intent lasts longer than AI done in panic.

14/10/2025

Why Agentic Workflows in Credit Underwriting Are a Different Beast?

When people talk about “agentic workflows,” they often imagine generic task orchestration — summarize this, extract that, call an API, move data.

But credit underwriting isn’t just another automation flow. It’s a reasoning system in disguise.

In underwriting, every “agent” must balance judgment, compliance, and data interpretation — not just ex*****on.

A loan file isn’t a single document; it’s a conversation between numbers, narratives, and policy.

That’s why:

Agents must reason, not just respond. A revenue mismatch or expense anomaly isn’t a failure — it’s a dialogue between human credit logic and data signals.

Context persistence matters. One borrower’s case can span dozens of inputs and weeks of iterations — true autonomy needs continuity.

Transparency beats speed. Decisions must be explainable, auditable, and regulatory-friendly — not black boxes.

When integrating BharosaAI + Finlens, our goal has been to make agents that underwrite like analysts, not like scripts — capable of conversation, validation, and convergence.

Agentic AI in credit isn’t about replacing analysts.

It’s about giving them a thinking partner that understands both numbers and nuance.

Hobbiate

30/03/2026

18/03/2026

14/03/2026

23/02/2026

09/02/2026

23/01/2026

23/12/2025

06/12/2025

17/11/2025

10/11/2025

03/11/2025

14/10/2025

Address

Telephone

Website

Alerts

Contact The Business

Shortcuts

Share

Category