📊 Offset: your personal AI CFO

The best consumer founders know the game. 26K+ of them read Consumer Startups every week.

Stay ahead. Get the playbook behind today's breakout startups.

Read time: 3 mins 23 seconds

Two years ago, Raj Khare was on top of the world.

He’d dropped out of Carnegie Mellon, built a personal AI computer that went viral, raised $4M from Naval Ravikant and other billionaires, and racked up six figures in pre-orders.

Then, almost overnight, he was forced out of his own company.

A year later, at 24, he’s back and building something equally ambitious: Offset, your personal AI CFO.

A few days ago, I bumped into Raj. Turns out, he’s been a longtime Consumer Startups reader. 

He told me his wild story, and I am excited to share it with y’all. 

Raj grew up in Mumbai with traditional Indian parents. Anything less than an A was unacceptable. He was following someone else's playbook… until someone gave him the book Zero to One for his birthday at age 15.

That book changed everything.

He taught himself to code. Websites, apps, games, anything he could build. He became obsessed with one thing: getting to Silicon Valley.

He figured the best path to get there was school. He transferred from a university in India to Carnegie Mellon during his junior year, but that didn’t last long. 

He dropped out immediately after getting accepted into ZFellows and subsequently HF0 with his friend, Srikanth.

P.S. HF0 is an incredible accelerator program, only 10 teams per batch, each getting up to $1M in investment. It produced three unicorns in its first year (Ramp, Pave, and Fabric).

Raj and Srikanth's first idea was to build a LLM fine-tuning as a service business. They grew it to $100K ARR, but they soon realized the margins were terrible. The only way to make it work is to raise huge amounts of cash and get to scale.

They decided to pivot the business to something more ambitious.

The thesis: While everyone was building AI in the cloud (OpenAI, Anthropic, etc.), there was a massive opportunity to go the opposite direction. What if AI ran locally? On your own hardware. Like Apple vs. IBM mainframes.

They called it Truffle, an AI computer that can run large AI models in your home with your personal hardware device.

They started building Truffle out of the Harvard SEC lab.

After a viral moment on Twitter, they brought in over $100K in pre-orders and raised more than $4M from Hummingbird, Naval Ravikant, Riaz Valani (the billionaire founder of Juul), and others. They soon outgrew the Harvard lab, started shipping units, and moved into a larger office in LA.

Then came the gut punch.

Raj and his cofounder started seeing the vision differently, over product strategy and execution. As a result, Raj had to step down. 

Despite the unfortunate co-founder breakup, Raj went back to the idea maze looking for another big problem to solve.

One key realization for Raj at the time was that AI hardware for consumer use cases might still be 3-5 years too early. AI models are bottlenecked by the memory of the device, not computing power. GPU memory is expensive for consumer use cases. He wanted to create something valuable in AI software this time.

The breakthrough came from a simple question:

Excel has 2 billion users. It hasn't changed in 40 years. What would it look like if you designed Excel from the ground up in the world of AI?

The whole economy can't function without spreadsheets. Every founder, every finance professional, every analyst—they all live in Excel.

Many people are already using AI tools like ChatGPT to supercharge their data analysis effort. In fact, 3.4% of all ChatGPT queries are related to this issue.

However, horizontal platforms like ChatGPT are inadequate for professional Excel users. 

Most professionals can’t afford to upload their sensitive financial models to ChatGPT. Compliance won't allow it. Security won't allow it. On top of that, many financial models are too HUGE and complex for those horizontal AI platforms to handle well.

Enter Offset.

Offset’s V1 is positioned as a juiced up version of the Excel AI copilot.

For banks and finance teams, Offset offers two critical benefits:

Better at giant Excel models: Deep integration with Excel means it understands complex formulas and nested logic. It can handle hundreds of thousands of rows better than ChatGPT or Claude. 

Privacy guarantee: Offset offers a local version, so sensitive financial data never leaves your computer. This is a compliance dream for banks that literally cannot use ChatGPT. His experience building AI personal computers @ Truffle prepared him well for this moment.

For Excel users, Offset is a powerful way to:

Offset isn't just an Excel copilot. Raj is looking to build your personal CFO.

He sees spreadsheets as an entry point into the game… overtime, he wants to build it like your own personal CFO that runs on your computer.

The 5-year vision: An AI that knows everything about your company (revenue, burn, runway, unit economics), advises you on high-leverage decisions (financing, hiring, expenses), and takes action through integrations with QuickBooks, Mercury, and banking.

You tell it to "cancel all unused subscriptions" → it actually does it on banking platforms like Mercury.

To be frank, Raj is up against some pretty tough shit. 

ChatGPT is free, everywhere, and "good enough" for most people. Long-term, OpenAI or Microsoft might build this into Excel natively as well. They have infinite resources. They own the distribution. On top of that, finance teams don't disappear easily. There's organizational resistance and political inertia. 

BUT here's why it could work: 

The privacy and compliance wedge is real. Big banks literally cannot use ChatGPT for this. Analysts at Morgan Stanley, Goldman, and any major financial institution can't upload their models to ChatGPT. But they can use Offset.

Raj has learned the timing lesson from Truffle. He won't be 5 years early again. The tech is ready now. Models are faster. The economics work. His local AI model expertise from Truffle days is rare. 

To succeed, the key is to find the one killer use case that makes people stick. Land one big bank as a design partner. And don't fall in love with the vision so much that he ignores what might bring value to users today.

–

I love founders like Raj.

Sure, him being a longtime reader of this newsletter helps 😉

More than that, Raj embodies the gritty, scrappy energy I admire most in founders.

He moved halfway across the world from Mumbai. Taught himself to code. Got into CMU and dropped out. Raised $4M from investors like Naval by going viral on Twitter. Built actual AI hardware and shipped it. Then had to step down from his own company. That's brutal. Most founders would've quit.

Raj is raising a $2–3M seed extension (after already bringing in $1M+ from angels and HF0 — again!). If you’re interested in the round or want to check out Offset, reach out to him and try the product for free here.

That's it for this week. See you next Tuesday.

- Leo

Follow me on X and LinkedIn

📥️ Want to advertise in Consumer Startups? Learn more.