Pujun Bhatnagar

With a focus and deep expertise in machine learning and all things AI, my journey has evolved into the role of Cofounder and CEO at Kintsugi, where I apply my technical prowess in Machine and Deep Learning to create vertical models that serve as indirect-tax engine that power Kintsugi AI. 

My tenure as a Senior Machine Learning Engineer at Facebook was marked by significant contributions to ML filtering and Ranking models, enhancing the precision and effectiveness of Fb, Instagram and Messenger Stories product. The skills honed there now underpin my strategic direction at Kintsugi, guiding a team in crafting cutting-edge, AI-driven contextual models.

Cofounder, CEO 

During my grad years at Stanford, Mark Zuckerberg and Facebook executives often showed up in the graduate CS courses. It was the perfect place to find students already deep in code and hungry to build. Snapchat was taking off on campus. Everyone in their 20s was using it, and it was the first product that genuinely challenged Facebook’s dominance.

After one of those lectures, he and my professor, who had become a mentor, were talking about Snapchat’s growth and what Facebook should do about it.
They needed people who understood the math behind ranking and modeling human behavior. 

My professor mentioned my name.

A few minutes later, I’m walking across Stanford’s campus with Mark, holding a beer (yes, the open-container policy is real), and he asks about my background. I told him what I’ve told people my whole life: math was the one thing that made sense to me when nothing else did. 

Machine learning felt like the natural extension of that, using data to understand what people like, what they don’t, how they behave, and why.

I talked about my work at Georgia Tech, the early AI research I was doing, and how much of machine learning is just applied statistics. And the whole time, I genuinely thought we were just chatting. It didn’t feel like an interview.

10 minutes into the walk, he said, “Why don’t you come help us get this started?” and convinced me to join. That was basically the entire hiring process.

It was right place, right time and that moment shaped the next nine years of my life. By the time I left Facebook in 2021, our team was ranking roughly 30 billion Instagram Story requests every single day, deciding what to show you, in what order, and why.

People sometimes ask why I’m working in something as “boring” as sales tax now. But the same part of me that loved that 10-minute conversation loves this work too. It’s all data, it’s all modeling, it’s all understanding systems people take for granted and finding the structure beneath the chaos.

My career has had its share of unexpected turning points. Meeting Mark Zuckerberg at Stanford set me on a path that eventually led to nine years at Facebook, where I became one of the early machine learning engineers working on ranking and filtering systems across Facebook, Instagram, and Messenger.

The work we did there shaped how billions of people consume content. It also shaped how I think about building products today.

Those years honed the instincts I now bring to Kintsugi as we build AI-driven compliance systems that are just as reliable and scalable as the models I helped ship at Meta.

Cofounder & CEO@Kintsugi: building AI infra OS for global indirect tax compliance | MIT & Harvard MBA dropout | CS@Stanford

Most days as a founder feels like getting punched in the face while you still have to make a hundred decisions anyway. I actually track this, between me and my co-founder Jeff, we make roughly 88 decisions a day. 

If I’m doing well, I get around 80% of them right. The other 20% kicks me in the teeth and humbles me fast.

You’ll get kicked in the face, you’ll be tired, your partner will be annoyed, your parents will wonder why you’re never around, and you’ll look at your bank account and realize you made significantly more money when you were at Big Tech.

And then you wake up and you do it again.

Founding a company rarely looks like a perfect strategy deck. That’s why I like tracking decisions. Part of it is just having a paper trail that a decision actually got made, instead of pretending an outcome “just happened.”

But a bigger part is forcing myself to look back and admit where I f****** up, and whether fixing it moved the needle or just moved the bottleneck somewhere else.

The right inner circle is incredibly important, which is why I mostly hang out with founders now. 

They don’t let you hide behind “on paper things are good.”  They’re the first people to give you a much needed objective perspective.

It’s easy to lie to yourself.

But, other founders help you identify bottlenecks, spot the thing you’re avoiding, ask what’s costing you sleep, and push you back toward making better decisions instead of just feeling better about the ones you already made.

Most competitive deals we see at Kintsugi right now are exits from Avalara. Teams are actively looking to leave a setup that stopped working for them.

The pattern is consistent:

- Support becomes unreachable at the exact moment filing pressure increases.
- Price increases arrive at renewal without any improvement in service.
- Integrations quietly degrade, especially after acquisitions, and teams lose confidence in whether the data feeding filings is actually correct.
- Nexus crossings and international thresholds are constantly missed.

By the time these teams talk to us, the decision to change has already formed. What they’re evaluating is risk: Will switching create new problems? Will the system hold up under pressure? Will someone be there when something goes wrong?

That’s why the questions sound straightforward but carry urgency. “How do you handle nexus?” usually means concern about getting visibility before crossing thresholds. “What’s pricing compared to Avalara?” reflects frustration with rising costs. “Can I talk to a human?” comes from past experiences where support disappeared when it was needed most.

January makes all of this more immediate. Filing deadlines turn frustration into action, conversations that stalled in November resurface with urgency.

This window won’t stay open forever, because service issues get addressed over time. But right now, dissatisfaction is widespread and visible, and teams are actively looking for a solution they can trust.

In tax, companies change systems when reliability breaks down. And once that happens, the decision tends to move quickly.

Cofounder & CEO @Kintsugi: building global AI powered Indirect Tax Compliance Infrastructure for the internet

If you’re a SaaS company with PMF and north of $2M in revenue, there’s a silent surprise waiting for you in your next round or acquisition.

And most founders don’t know what it is, because there’s still an education problem. But due diligence forces a simple question to the surface: are you tax compliant everywhere you have customers?

If the answer isn’t clear, everything slows down.

Now you’re budgeting time, money, people, and leadership attention just to get back to neutral. At Series B and beyond, it’s even more painful. The stakes are higher, expectations are tighter, and the window to fix it is smaller.

On the e-commerce side, the failure mode looks different but the outcome is the same. Founders know sales tax exists, so they throw money at the problem. They trust a ‘trust-me-bro’ solution.

When you actually look under the hood, there are no registrations happening. No filings happening. Nothing is actually getting done.

When this shows up, it doesn’t show up politely. It shows up urgently, and it directly affects your growth plans.

For SaaS, this risk tends to surface once you’re post-product market fit and scaling, usually around $2M and above. For e-commerce, revenue barely matters. One viral channel or one TikTok spike can trigger compliance overnight.

None of this is random. 

Sales tax kills deals because teams assume it doesn’t apply until someone else forces the conversation.

I didn’t realize how hard it is to hire people who will get their hands dirty until we scaled. Everyone wants to join a fast-growing startup, everyone says they’re hungry, everyone says they want to build. And I believe most of them when they say it. The disconnect shows up when the work starts:

At Kintsugi, marketing means being in front of customers constantly. Sitting in demos with finance teams. Chasing down users for case studies while they’re in the middle of audits, filings, or expansion planning. Iterating on messaging that doesn’t land yet because the problem we’re solving only becomes obvious once something breaks.

Customer success means jumping into escalations when filings don’t look right or a notice shows up. Owning the outcome even when the issue touches product, data, and external systems all at once. Being close enough to the customer that you feel the friction in real time, not through a summary.

Sales means staying close to the truth. Hearing why a CFO is nervous, why a deal is stuck in diligence, why “we’ll handle this later” keeps coming up. Feeding that reality back into the product instead of smoothing it over in a forecast.

This gets harder when hiring from larger companies. A lot of people are excellent at slides, alignment meetings, and theoretical strategy. That experience can look impressive on paper. 

But early and mid-stage companies like ours don’t move forward because of well-phrased decks. They move because someone is willing to sit with messy data, incomplete systems, and real customers until the problem is actually solved.

What I’ve learned is that the real signal is whether someone is willing to do the work that directly touches customers and product, even when it’s uncomfortable or unpolished.

If someone can’t get into the weeds, they won’t scale with the company. And you don’t learn that from a resume or an interview loop.

You learn it once the work starts.

If I were a Controller at a $5-10M SaaS company, I would spend most of my time trying to reduce uncertainty.

At that stage, the business is growing faster than its systems. New customers, new states, new pricing models, new tools. Revenue is finally material, but the infrastructure behind it is still uneven and often built for where the company was, not where it’s going. 

Here’s how I’d reduce risk without slowing the business down:

1. Establish a clear picture of my current tax exposure.

My first priority would be understanding where the company already has tax exposure. That means knowing which states we’ve crossed economic nexus thresholds in, which products or services are taxable by state, and where exemptions are being handled manually or inconsistently. I would also want to know which filings rely on reminders, spreadsheets, or individual knowledge. Until this is clear, it’s impossible to separate real risk from noise.

2. Surface where risk accumulates quietly.

Most issues at this stage build slowly as volume increases and processes drift. The first real signal is usually external, like a state notice, a diligence request, or an investor question that should be easy to answer, but isn’t. Anywhere the answer depends on assumptions rather than data is a place where risk is already forming.

3. Be deliberate about automation.

I would be cautious about automating too early. Early decisions have a habit of becoming permanent, even when they were made under time pressure. Before adding complexity, I’d focus on clean transaction data, clear ownership of tax decisions, and systems that make it easy to trace how numbers were calculated. Those foundations tend to scale better than layered workarounds.

4. Preserve flexibility as the company grows.

At $1-5M in ARR, I wouldn’t want long commitments or rigid systems. I’d look for tools that provide visibility across states early, allow exposure to be addressed gradually, and make it easy to add automation as volume increases. Flexibility matters more than completeness at this stage because the business model is still evolving.

At this stage, a controller’s job is to keep the company moving forward without accumulating hidden risk that shows up later, when the stakes are much higher.

taplio

Pujun Bhatnagar's Linkedin Analytics

Pujun Bhatnagar

Performance ScorePujun Bhatnagar's LinkedIn Performance

What is Pujun talking about?

Pujun Bhatnagar's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Pujun Bhatnagar's Linkedin Analytics

Pujun Bhatnagar

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Pujun Bhatnagar's LinkedIn Performance

What is Pujun talking about?

Pujun Bhatnagar's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Pujun Bhatnagar's Linkedin Analytics

Performance ScorePujun Bhatnagar's LinkedIn Performance

Pujun Bhatnagar's Best Posts (last 30 days)