Best AI Startups in NYC to Work at Right Now

New York has crossed a line that's easy to miss if you're only watching the funding headlines out of San Francisco. The city now builds frontier research labs, AI infrastructure companies, and world-model developers, on top of the applied-AI strength it already had in finance, media, healthcare, and enterprise software. The old story was simple: San Francisco builds the models, New York puts them to work in banks and ad agencies. That story doesn't hold anymore. Companies like Reflection AI and Flourish are doing foundational research work in New York, not just building a layer on top of someone else's model.

The numbers back this up. NYC Tech Journal counted more than 2,000 AI startups in the city as of August 2026, backed by a large pool of workers with AI skills and a deep bench of active venture capital firms. That's the infrastructure of a real hub, not a secondary market riding on another region's coattails. The growth shows up in physical space, too: AI companies leased a substantial amount of Manhattan office space in 2025 alone. That's a spatial cluster forming, people showing up to buildings every day, not a cluster of mailing addresses on pitch decks.

The talent pipeline feeding all this is local. Columbia, Cornell Tech, CUNY, and NYU have turned out tens of thousands of graduates with AI-ready degrees in recent years, and that number compounds every year a new class graduates into the same fifteen-mile radius. With the research labs, the infrastructure, the office space, and the graduates all in place, an engineer no longer has to ask whether New York has real AI companies worth working at. It has plenty. Out of thousands of companies, the real question is which ones are worth your time right now.

How this list was built

Funding size and valuation make for bad shortcuts when you're deciding where to spend the next few years of your career. A company sitting on a massive round with no paying customers is a fundamentally different bet than one with strong annual recurring revenue and a fresh round behind it. Both situations can be worth joining, but they carry different risk, and treating them as the same signal will lead you wrong. NYC Tech Journal built its list around four criteria: real New York roots, AI that sits at the center of the product rather than bolted on as a feature, evidence of actual technical or commercial progress, and a credible reason the company matters more in a few years than it does today. This piece applies that same filter, with an engineer's priorities layered on top: team size relative to ambition, since a small team on a big problem means more individual leverage, whether the role is in-office or remote-native, and whether the hiring process moves fast enough to be worth your time starting it. The companies below are grouped by the kind of engineering bet each one represents, frontier model research, generative and creative AI, enterprise document intelligence, vertical AI in specific industries, because the right choice depends on what you want to build, not which logo has the biggest round attached to it.

Frontier model bets: Reflection AI and Flourish

New York now has two credible entries in the race to build frontier models, a sentence that would have sounded like wishful thinking two years ago. Reflection AI has raised more than $4.6 billion in reported funding, making it one of the most heavily capitalized AI startups in the city by a wide margin. Reflection AI's open roles include Member of Technical Staff positions in Pre-Training, Post-Training, Evaluations, and Safety, along with Forward Deployed Engineer roles focused on LLM Post-training. The Member of Technical Staff roles are listed across both New York and London, while the Forward Deployed Engineer roles in Post-training are listed in New York and San Francisco only. The listings split across three locations rather than sitting in one place, so confirm the specific city tied to a role before you get far into the process.

Flourish represents a different research paradigm rather than an incremental improvement on the architectures everyone already knows. That makes it a higher-variance bet, but also a higher-ceiling one for engineers who want to work on something genuinely new. The obvious objection to betting on frontier research in New York is that you're competing for talent against San Francisco labs with a decade-plus head start. Fair enough, but both Reflection AI and Flourish have shown they can raise capital at frontier scale, and that's the first real test of whether a lab is serious or just well-marketed. Nobody can tell you today which lab wins the frontier race. Nobody honest would try. If you want to be in the room where foundational model work is actually happening outside San Francisco, these two are the rooms that exist in New York right now.

Generative and creative AI: Runway

Runway marks the shift from research bets to a company that has already shipped a real product at real scale. Where Reflection AI and Flourish are still proving the model layer works, Runway has spent years turning generative video into commercial infrastructure that studios and creative teams actually use. The company has also outgrown its original label. Runway started as an AI video tool and has moved toward building world models, and that shift has widened the engineering surface area well past creative tooling into the model architecture itself.

Runway was founded in 2018 and is headquartered in New York City, but it runs as a remote-native company with strong async practices, and many of its engineering and research roles are open to remote candidates. That's rare on this list, and it matters: Runway is one of the few NYC-headquartered AI companies here where living outside Manhattan isn't a disadvantage. The team is small relative to the size of what it's attempting. Engineers joining now often work directly alongside the people who wrote the research papers the product is built on. That's a real learning and leverage signal, not a recruiting line. As the company's ambitions expand from creative tools toward broader world models, the work expands with it, producing engineers who aren't just building better video generation but working on the architecture underneath it.

Enterprise document intelligence: Hebbia and AlphaSense

This tier moves from the model layer and creative tooling into enterprise knowledge work, where the problem is making AI genuinely useful for people who read and synthesize documents all day for a living. Hebbia and AlphaSense are both attacking that problem, but from different stages and at different scales, and the contrast between them tells you something about what "winning a workflow" actually looks like. The most durable companies in this space will be the ones that own a specific high-value workflow start to finish, rather than adding an AI feature to an existing piece of software. Hebbia and AlphaSense sit at two different points on that path.

Hebbia was founded in 2020 by George Sivulka and is backed by Peter Thiel and Andreessen Horowitz. It raised a sizable Series B, and its Matrix 2.0 product launched in 2026. The company powers investment decisions for BlackRock, KKR, Carlyle, and Centerview, among a large share of the world's biggest asset managers, putting it in direct contact with firms that manage enormous pools of capital. Current open roles include Software Engineer for Infrastructure, Backend Engineer for the Agent Collaboration Platform, Forward Deployed Engineer, and AI Strategist for Corporate Law, and all of them are in-office. Hebbia's culture leans hard into in-person collaboration, so engineers who want remote flexibility should know that going in. The Agent Collaboration Platform role points toward multi-agent orchestration for professional research tasks, a step past simple single-query retrieval. The use case, helping financial and corporate analysts search and synthesize research, is one where a wrong answer has real financial consequences, so the bar for reliability and evaluation engineering is genuinely high here.

AlphaSense sits further along that same curve. It has built past half a billion dollars in annual recurring revenue, which puts it in a different risk category than an earlier-stage company still proving its model works. That's real revenue at scale, not a growth story funded purely by the next round. For an engineer weighing the two, Hebbia offers earlier-stage upside with an in-office culture, while AlphaSense offers the lower-risk profile that comes with large, proven revenue. Neither is the better choice in the abstract. It depends on how much risk you want attached to your next few years.

Vertical AI in housing, healthcare, and finance: EliseAI, Rogo, Norm AI, and Tennr

This group has each picked one industry, gone deep into it, and started winning there, and together they make the clearest case that New York's industry density, real estate, finance, healthcare, law, is a structural advantage for AI startups rather than a historical accident the city happens to benefit from. The vertical AI companies most worth joining are the ones operating in industries specialized enough that a generic model can't compete on its own, and New York's concentration of exactly those industries means the engineers building these products can get customer feedback from two blocks away.

EliseAI is hiring at a scale that stands out even on this list, with 122 to 124 active roles open across engineering, sales, strategy and operations, product solutions, and customer success. Like Hebbia, its culture is built around in-person collaboration in New York. Backed by a16z and running strong ARR alongside more than a hundred open roles, EliseAI is clearly in scale-up mode rather than its founding stage, and the nature of the engineering work has shifted accordingly. The company already serves 28 of the top 30 property owners in its market, so the problem engineers join to solve isn't "does this work," it's "how do we keep this working reliably at the scale we're already operating at."

Rogo works in financial services, a space where regulatory and accuracy constraints run high by default. The engineering challenge there has less to do with getting a model to produce an answer and much more to do with proving that answer is reliable enough to act on in a high-stakes financial workflow.

Norm AI raised a nine-figure Series C at a unicorn valuation and builds AI agents focused on law, regulation, and compliance. In that domain, a hallucinated answer isn't a minor UX complaint, it's a liability with legal weight behind it. The engineering problem is building agents trustworthy enough to operate inside adversarial regulatory environments, where being wrong carries a real cost.

Tennr raised a significant Series C and builds purpose-built models to fix specialist referral workflows in healthcare, an area still choked by faxes, scanned PDFs, and phone calls passed between offices. That fragmentation makes healthcare referrals a high-value target for AI that can read messy, unstructured documents and route them correctly, and it's the kind of problem that rewards an engineer who wants to see a model's output change a real-world process the same day it ships.