On September 9th, Y Combinator will host its 43rd Demo Day, where more than 150 companies will present to an audience of investors and media. The event marks the culmination of YC’s three-month program, during which founders receive funding, intensive mentorship, and access to a powerful network of peers and investors. While the structured program lasts only a few months, support from the YC community extends far beyond.
Meet The Builders
Every YC batch is unique, and Summer 2025 is no exception. The level of grind, energy, and ambition has been extraordinary—pushing the boundaries of what most would think possible. And while the diversity of ideas is striking, it’s no surprise that the majority of startups this time are building in AI, reflecting both the urgency and the opportunity of the moment.
Luminal compiles AI models to deliver the fastest, highest-throughput inference cloud in the world. In the race to get to market first, most AI companies are leaving thousands of dollars on the table each month by failing to optimize their models. Luminal solves this by making AI models run up to 10x faster through its purpose-built compiler and cloud platform.
Today, builders often stitch together tools like Lovable, n8n, and Supabase just to get a basic app running. These ‘Frankenstein’ setups are fragile, insecure, and rarely production-ready. Non-technical users still lack a single solution that can take them from idea to full-stack app. VibeFlow solves this by unifying UI, backend, and database into one platform—enabling anyone to go from prompt to production in seconds.
Enterprises are racing to adopt AI tools, but most have little visibility into which ones employees are using—let alone whether they’re secure or compliant. The result is shadow AI, compliance risks, and wasted spend. Tecto addresses this with an ‘HR system for AI employees’ that discovers, vets, and continuously monitors third-party AI tools. With regulations tightening and AI adoption accelerating, the timing couldn’t be more urgent.
Every year, Brazil’s digital workers lose billions in hidden taxes and fees just to access money they’ve already earned. Cacao unifies what once took three or four apps into a single platform—letting users receive USD, save in stablecoins, transfer BRL instantly via Pix, and spend globally with a Visa Platinum card. With Pix adoption surging and stablecoin payments newly legalized, now is the inflection point to make cross-border money instant, cheaper, and compliant.
The Prompting Company helps products get discovered inside ChatGPT and other large language models. As platforms like ChatGPT, Perplexity, Claude, and Gemini become the new starting point for brand discovery— with chatbot traffic more than doubling in the past year—the shift is undeniable. The transition is happening fast, and investing in AI-driven visibility today gives brands a critical edge over competitors.
April is a voice AI executive assistant that manages email and calendars hands-free—built for busy professionals who need an executive assistant but don’t have one. With voice AI now achieving sub-second latency, truly conversational interactions are possible. Advances like MCPs and tool calls are making AI systems more reliable and action-oriented, enabling them to perform meaningful tasks rather than just generate text.
Software is shifting from static apps to agent-driven systems powered by MCP. The challenge is that dev teams want to build agents, but end up wasting time wrestling with server configs, authentication, and fragmented infrastructure—nobody wants to reinvent the plumbing. This company solves the problem with a full vertical stack, from an SDK that makes MCP agents simple to build to a cloud control plane that manages hosting, security, and observability.
Companies are capturing more video than ever, but analyzing it with traditional computer vision remains slow, costly, and inflexible. Just as LLMs like ChatGPT unlocked reasoning over text, a new class of AI called Vision-Language Models (VLMs) can now do the same for visual data. OnDeck’s task-specialized platform allows businesses to ask complex questions of their footage and get answers instantly—without months of engineering. We’re at the dawn of a new era for perception, and OnDeck is the infrastructure layer making this power accessible and scalable for enterprises.
AI is creating a generational opportunity to reimagine how music is made. Riff is a music editor with superpowers—helping artists work faster and making it easier than ever for hobbyists to break into music.
Design Arena is a forcing function for AI design and visuals. Today, models can win IMO gold, yet still generate something as flawed as white text on a white background. Design Arena benchmarks models on their visual quality across websites, game design, 3D assets, and more—while also putting builder tools like Lovable, Bolt, and v0 head-to-head.
Verne Robotics builds AI models that teach robots new skills in hours—cutting training times from months to mere hours. This rapid learning makes robot automation far more accessible to companies of all sizes. Verne’s robots help businesses cut costs by automating back-of-house tasks, powered by an innovative pay-by-the-hour business model.
Physical technologies like autonomous vehicles, robotics, and intelligent devices are set for explosive growth in the next decade, but progress is bottlenecked by the firmware that powers them. Embedder is building an AI platform that enables engineers to create, test, and deploy firmware across complex embedded systems in record time.
Minimal is building AI customer support teams for e-commerce companies. Traditional chatbots frustrate customers, and existing AI agent solutions are complex to set up and manage. Minimal makes it simple for brands to create and manage AI support agents that automate up to 80% of customer interactions across chat, email, social media, and WhatsApp. At the core is Minimal’s AI Manager—an agent that manages the agents. Instead of fiddling with dashboards or writing prompts, users can simply talk to the AI Manager, making setup as easy as talking to a colleague.
Uplift builds foundational voice models for underserved languages. Developers use its APIs to create voice-first digital experiences—from e-commerce and banking to education—unlocking opportunities for millions of users who prefer voice over text.
Hospitals are rapidly adopting AI tools—from scribes and imaging models to triage bots—but lack the infrastructure to evaluate or monitor them. A single biased or drifting model can harm patients, trigger lawsuits, and violate new regulations like ONC HTI-1, California SB 3030, and the Colorado AI Act. The challenge isn’t willingness—hospitals want AI—but the inability to deploy it safely at scale. Without governance, most projects stall, leaving hospitals stuck in pilot hell. Parachute enables hospitals to adopt AI safely and at scale by vetting tools that fill clinical gaps, running state-of-the-art evaluations, monitoring every model for drift and compliance, and recording every decision in an immutable audit trail.
The Stories Behind the Ideas
Every startup begins with a story. Behind the relentless grind of building and fundraising are founders shaped by personal struggles, breakthrough insights, and defining life experiences. What makes them extraordinary are the unique journeys that led each team to take the leap into entrepreneurship—and to reimagine what’s possible.
For some, experience at big tech (Intel, Amazon and Apple) was the foundation. Each brought something unique to the table at Luminal: Joe Fioti worked on machine learning hardware, Matthew Gunton worked on global scale cloud deployments and Jake Stevens grew his previous startup to $5M in ARR.
For some, the spark came from watching consumer behavior change in real time. After co-founding Typedream (YC W20), a website builder later acquired by beehiiv, Michelle Marcelline realized that the SEO playbook was becoming outdated. “People are asking AI instead of searching,” she explained. That shift inspired The Prompting Company, built to ensure products get discovered inside AI answers as fewer people rely on search engines.
Others built solutions to give non-technical creators the same freedom and control developers enjoy, without writing a single line of code. Alessia Paccagnella and Elia Saquand, co-founders of VibeFlow, knew how powerful AI coding assistants can feel for non-technical users. “For developers like us, it’s easy—we can debug and refine what AI generates. For non-technical users, AI feels like a black box,” she said. With VibeFlow, you have visual tools like a backend editor and visual Git, so users can watch AI change their app, debug issues, and guide the system with better context.
For Neha Suresh, the pain point was as relatable as an overflowing inbox. Before starting April, she and her co-founder Akash Thakur were engineers overwhelmed by back-to-back meetings and endless email. “We needed an EA and couldn’t afford one, so we built one the way we know best—using technology,” she said. The result: a voice AI executive assistant that manages email and scheduling hands-free, designed for busy professionals like themselves.
Some founders were drawn to emerging protocols with world-shifting potential. “Since the first release of ChatGPT, we’ve been obsessed with how agents connect to external systems,” said Luigi Pederzani, co-founder of mcp-use. When the MCP protocol launched, they saw its power but also its limitations, locked inside proprietary apps like Cursor or Claude Desktop. Their mission became clear: unlock MCP for everyone, giving developers the ability to build and deploy agentic applications anywhere.
For Sepand Dyanatka and his co-founder Alexander Dungate, frustration turned into opportunity. “We were our own first, and most frustrated, customer”, he says. In his last company, computer vision hit a wall: too little training data, and models that simply failed to generalize. That pain point inspired a pivot to Vision-Language Models. The engine they built solved their problem instantly—and today powers OnDeck AI, enabling industries from defense to infrastructure to analyze video with unprecedented accuracy.
Co-founders Niosha Afsharikia and Roksana Baleshzar bring together deep expertise in AI and enterprise governance. Roksana has spent her career evaluating and deploying AI at scale, while Niosha has led major consulting projects in compliance-heavy industries. Together, they bridge both worlds—building cutting-edge AI systems and implementing them responsibly inside complex organizations.
And in other cases, the bond was friendship. The co-founders of Parachute, Aria Vikram and Tony Yamin began as close friends trading ideas between cutting-edge AI and the real constraints of healthcare. That friendship turned into a partnership built on trust and complementary skills. Together they decided to build Parachute, creating the governance layer that can bridge innovation and compliance inside hospitals.
Neil Nie and Aditya Jha, co-founders of Verne Robotics, met in high school building robots. “Neil built a self-driving golf cart; I tinkered with delivery bots,” Aditya said. “Robots that learn and act alongside humans have always been our passion.” Today, Verne trains robots in hours instead of months, making automation accessible to any company.
Embedder is building a software product to solve a hardware problem. Ethan Gibbs brings experience in AI and distributed systems, while Bob Wei has expertise in firmware and robotics, which gives them a unique edge. The space is predominantly occupied by highly specialized engineers who lack the cross-domain knowledge required to build developer tools.
For others, the motivation was personal creativity. “As music producers for over six years, this is the most personal problem we’ll ever solve,” said Adith Reddi, co-founder of Riff, an AI-powered music editor. “Current tools are clunky and outdated, and we saw this as the golden moment to reimagine music creation.” Fresh out of UC Berkeley, he, together with his co-founder, Jinlin (Kenny) Zhang, committed to building full-time.
And sometimes, the story is much more personal than technical. “I worked on voice tech at Alexa and Siri for about nine years,” said Hammad Malik, founder of Uplift. “But Big Tech was moving too slowly. It was also personal—I had never been able to talk to voice models in my own language until we created our own.” Uplift builds foundational voice models for underserved languages, unlocking digital experiences for millions of users worldwide.
The sky is the limit, and YC is the perfect launchpad
What happens after the three-month sprint ends? YC’s impact doesn’t stop with Demo Day—it reshapes how founders think, build, and dream. Ask a dozen companies and you’ll hear different answers, but the throughline is clear: the program leaves a lasting, transformative mark on every team that goes through it.
“YC level-sets what ‘normal’ can be,” explains Neil Nie, co-founders of Verne Robotics. And that new definition of “normal” is often far beyond a founder’s expectations. “We’ve 10xed our ambitions since starting YC,” adds Luigi Pederzani, co-founder of mcp-use.
The program doesn’t just accelerate progress—it forces founders to think bigger. Alessia Paccagnella offers a clear example: “At first, we focused on helping people launch micro-SaaS apps and internal tools faster. But through the batch, we saw the larger potential: VibeFlow can become the ‘CTO’ for non-technical founders. The need is massive, founders without technical backgrounds are underserved, and we believe some of the biggest companies of the next decade will be built on VibeFlow.”
That same shift in perspective played out for Sepand Dyanatka, co-founder of OnDeck AI: “We entered YC aiming to dominate one vertical; we’re leaving with the conviction to power the entire visual analysis market.”
Adith Reddi, co-founder of Riff, calls it “the world’s most competitive bootcamp” for good reason. “The sense of urgency is so infectious among founders that it becomes the most productive period in most people’s lives.”
The co-founders of Cacao—Carlos Jimenez, Alec Howard, and Michael Mason—say the experience sharpened their focus. “YC forced us to distill our story and focus on what truly matters: growth, users, and clarity of vision,” they explained. Carlos adds, “The connections are unparalleled—doors to investors, operators, and alumni open instantly in ways no LATAM startup could achieve alone.”
However, it’s not only about intensity, one of the most important things is a community. Jake Stevens, co-founder of Luminal, puts it simply: “It’s absolutely the community. Everyone here knows we have a long journey ahead and is extremely supportive of helping each other get there. I’m paying for three of my friends’ products right now.”
That community is more than camaraderie—it’s a lifelong network. Niosha Afsharikia and Roksana Baleshzar, co-founders of TectoAI, note that the network can open doors in days that would normally take months, while surrounding you with fearless founders who push you to think bigger and move faster. Michelle Marcelline, co-founder of The Prompting Company, agrees: “The YC badge gives instant credibility, and with 300+ peers and experienced group partners, warm introductions to investors and customers are always one message away.”
Niek Hogenboom, co-founder of Minimal AI, summarizes it best: “It’s two things: learning from the best people in the world—YC partners, fellow batchmates, and alumni. Secondly, it’s realizing that building a unicorn is actually realistic. YC constantly invites alumni who have done it, and they were in our position just a few batches ago. The sky is the limit, and YC is the perfect launchpad.”
As YC’s 43rd Demo Day approaches, it’s clear that founders are not just pitching products—they’re rewriting playbooks, reimagining industries, and proving that bold ideas backed by relentless execution can scale faster than ever before. For some, it’s about solving personal pain points; for others, it’s about chasing opportunities that only this moment in technology makes possible. But together, they embody the same truth: YC is more than a three-month accelerator. It expands vision, compresses time, and equips founders with the confidence—and the community—to build the next generation of global companies.