The conventionally celebrated path follows that you get an internship, earn a return offer, and get a stable job. Harshil got multiple full-time offers. He turned them all down to go back to school.

Sometimes, the best return on an internship isn’t the job offer. It’s clarity about what you want to do next.

Over the past year, Reach has been meeting hackathon winners, speaking with students, investing in some, and helping others land at portfolio companies. What we keep seeing: the most interesting people don't fit neatly into a box. Researchers who build. Engineers designing with taste. Designers who ship daily code. People who do all of the above.

This community is full of them. So every month, we're sharing what building actually looks like at startups, schools, and everywhere in between. Today, we're starting with one of our own readers: Harshil V.

In today’s issue:

Read time: 5 minutes

He started doing research at Rutgers before he'd taken many CS courses. He read papers obsessively and went straight to the authors: asking thoughtful questions, showing genuine interest in conversations. A six-month project with Amazon, and the mentor who came with it, gave him the foundation to take it further. He fell in love with how machine learning works, not just the theory but the application too. Since then, he’s led his own independent research, presented it at conferences and published solo-authored papers on natural language processing.

“The best way to stay up-to-date on AI and machine learning is to read papers,” says Harshil. “Reading papers is the best way to know what researchers are coming up with in the field, and if that’s too difficult, a lot of people also write blogs, a slightly easier way of understanding cutting edge developments.”

Last summer, he spent five months at a Reach portfolio company with a vague brief: build a Chrome extension that creates resources in Google Docs. No instructions. No manager. Just a direction and a deadline.

What he didn't know yet was that Google actively tries to prevent exactly this. Programmatically inserting text into Google Docs (to make it behave like a human is typing) is something Google has been patching out for years. Maybe three or four extensions in the entire Chrome Web Store can actually do it.

Harshil figured out how to be one of them.

"I was pretty much reverse engineering their code," he told us. "Every time a method came out, Google would fix it. So I had to keep finding new ones."

It took months. But he got there. Hundreds of users are now using what he built.

He also interned at one of the largest retailers in the US to see what big company life was like, and discovered it wasn't for him.

Looking back at both experiences, ”the startup environment was totally different,” he recalls. “Total ownership. No one to pass the blame to. Wins and failures that land squarely on you." 

Internships and work experience are valuable — not just when they lead to full-time offers, but when they give you the conviction that the next best path is sometimes going back to school. In Harshil’s case, what he’s learned is that the most valuable thing an internship can give you is clarity in what you are truly excited about.

"Before this year, I had no idea what I wanted to do," he said. "Now I know I want to do something related to startups or research. That's more than I had before."

Harshil has been accepted at Carnegie Mellon and Harvard, and is waiting to hear back from Stanford. He’s turned down multiple return offers (at the retailer and at startups) to pursue his Masters in CS and Machine Learning. Allowing him to build technical range, continue research, and expand his network.

The most valuable thing you can build right now isn't a resume. It's a track record of finding hard problems and solving them in ways that are uniquely yours — using your background, your curiosity, and whatever tools you can get your hands on. Harshil didn't take the traditional SWE path. He had research instincts, a hard problem, and the resourcefulness to figure it out. 

If you’d like a chance to work with one of our portfolio companies, add your profile in the Student Repo Directory if you haven’t already. Founders looking through this curated list on a rolling basis.

Next month we’re chatting with Jared Silver, engineering leader at Brilliant, which is an interactive learning platform for math, AI, data, and CS. He’ll be sharing his take on opportunities for early-career engineers in an AI-enabled world. Questions for Jared? Send them to [email protected].

Can you build a venture-scale business selling $100 SaaS to pizzerias?

Most investors say no. The critique is familiar: small deals, fragmented customers, messy distribution. And if the market is limited to pizzerias already buying software, the skepticism makes sense.

But that assumption misses something fundamental. When technology makes a resource dramatically cheaper, total consumption goes up, not down. Lower costs don't just serve existing buyers more efficiently. They pull in entirely new ones, people who previously bought nothing at all.

That's what's happening now. AI is collapsing the gap between what a billion-dollar company and a local pizzeria can actually execute. Expertise that once required a $10,000 retainer (think: financial guidance, marketing strategy, operational infrastructure) is now available on demand at a fraction of the cost. The SMB market was never small. It just didn’t have the right tools.

Build the outcome, not the interface. The most valuable products here won't be dashboards or integrations. They'll be systems where the AI does the actual work. I.e., an agent that doesn't just surface financial data but tells the owner what to do about it. One that explores and generates ad creative, while also executing the entire campaign. The product and the service are the same thing.

Design for trust before you design for efficiency. A small business owner handing sensitive financial or customer data to an AI agent is a leap of faith. The builders who win here will focus on transparency, explainability, and user control — giving owners visibility into what the agent is doing and why, with clear moments to step in. Trust will be the determining factor for user retention.

Rethink how you sell and support it. Enterprise playbooks don't work here. Small business owners don't respond to cold outreach, long sales cycles, or complex onboarding. The best distribution models will be frictionless to start, embedded in workflows they already use, heavily localized and supported in ways that feel more like a helpful neighbor than a customer success team.

Our full thesis: https://www.reachcapital.com/resources/thought-leadership/ai-smb-economy/ 

Happy building, 

Valentina & the Reach Capital team

Reach Capital invests in early-stage founders redefining how we learn, live, and work. Our portfolio of 130+ startups includes tools you might’ve used in school (ClassDojo, Desmos, Brilliant) and next-gen, AI-native disruptors shaping how future generations build, work, and thrive (like Replit and GPTZero). 

These teams are constantly on the lookout for talented builders. Share what you’re studying, building, or exploring, and we’ll intro you to projects, people, and paid opportunities for you to build upon.