Most AI Video Startups Are Hiring to Find a Market. Mirage Is Hiring Because It Already Has 20 Million Users and $28.4 Million in Revenue.

The $75M Signal: From App to AI Lab

Mirage raised $75 million in growth financing on March 24, 2026, in a round led by General Catalyst's Customer Value Fund, according to TechCrunch. The deal brings the New York–based company's total funding past $175 million and marks a deliberate shift in identity: Mirage is no longer positioning itself as a consumer video-editing app. It's an AI lab.

The rebrand from Captions to Mirage, which took effect over the past year, was cosmetic on the surface but strategic underneath. The company trained its own model for pacing, framing, and attention dynamics in short-form video. It launched an audio model designed to preserve regional accents rather than flatten them into American English. CEO Gaurav Misra said the next set of models will focus on what he calls "assembly intelligence," stitching together video from multiple sources and components into a finished product. The Captions app, with its 3.2 million downloads over the past year and $28.4 million in in-app revenue, is now one output of a broader foundation-model research effort, not the whole company.

General Catalyst's CVF vehicle is non-dilutive and specifically designed to fund customer acquisition, which tells you something about how Mirage's pitch landed. Pranav Singhvi, managing director of the CVF fund, said Mirage's "business equation is extremely figured out" and that the company knows "exactly how to spend that dollar and generate a very attractive ROI." Andrew Ziperski, a partner at the same fund, called Mirage's customer acquisition engine "powerful" in a press release. The investor thesis isn't speculative. It's anchored in unit economics the company has already demonstrated.

The capital has two stated purposes: expand into high-growth Asian markets and invest in product innovation and foundation model development. Misra framed the broader logic in the company's blog post: AI has compressed the time it takes to build software features from months to weeks, eroding the traditional moat of product quality alone. "Distribution and capital efficiency are the new moat," he wrote. The $75 million is fuel for exactly that race.

What Mirage Is Actually Hiring For

Mirage's open positions page lists 13 active roles, all based at the company's Union Square headquarters. Five of those roles were added in the past week alone, according to Zero G Talent's board data — a pace that signals the $75 million round is already converting into headcount.

The engineering slate breaks into three distinct layers. At the foundation-model level, two ML engineer roles sit at the core: ML Engineer, Generative Video and ML Engineer, Agentic Systems. The generative-video posting calls for someone who can train and optimize large-scale video and multimodal models, implement distillation and quantization to accelerate diffusion and autoregressive generation, and build distributed training systems with an eye on GPU utilization and throughput. The required stack is specific: PyTorch, CUDA, Triton, and FSDP-level distributed training. The agentic-systems role signals Mirage is building orchestration layers — systems that can do the same assembly work, which Misra has described as "assembly intelligence."

Above the model layer, three software-engineering roles handle the product and infrastructure stack: Software Engineer, Backend; Software Engineer, Web Product; and Software Engineer, iOS. A fourth, Software Engineer, Agents, bridges the ML and product sides, building the agentic interfaces that let users direct video generation through natural language. There's also an Early Career software-engineering role, suggesting Mirage is willing to train junior hires rather than competing exclusively for senior talent.

On the product side, two Product Designer positions (one early-career, one open-level) round out the technical slate. The marketing and operations roles (three marketing positions and one technical recruiter) are smaller in number but reflect the freemium-to-enterprise growth strategy Mirage adopted when it switched its pricing model in January 2025.

Those figures put Mirage in direct competition with well-funded peers for ML talent in New York, and the requirement that all roles be in-person at Union Square narrows the aperture further.

What's missing from the list is just as revealing. There are no dedicated data-labeling roles, no prompt-engineering positions, and no research-scientist titles separate from the ML-engineer track. Mirage appears to want people who can move from prototype to production without handing off to a separate team, a staffing model that compresses the research-to-product pipeline and demands engineers comfortable operating across that full stack.

A Hiring Push That Responds to a Current Market

Before the General Catalyst–led round, Mirage had already accumulated 20 million-plus users who had created more than 200 million videos through its consumer app. Appfigures data shows that app generated $28.4 million in in-app revenue. Those aren't projections. They're usage figures from a product that existed before Mirage decided to build foundation models.

The distinction matters. Most AI video companies are hiring to build a market. Mirage is hiring because it already has one and needs infrastructure to match it. The five roles added to Zero G Talent's board in the past week — spanning iOS, backend, agentic systems, generative video ML, and performance marketing — map directly onto the gaps between where Mirage's user base is and where its foundation-model ambitions sit. A company speculating on product-market fit doesn't need a Senior Performance Marketing Manager alongside two ML Engineer roles at $175,000–$275,000 a year. A company monetizing 20 million users does.

The $75 million round funds the lab. The $28.4 million in revenue funds the hiring. That's the harder signal.

Why Generative-Media Engineering Is a Distinct Talent Category

The roles Mirage is filling don't sit neatly in any existing job taxonomy. Scan the titles on Zero G Talent's board — ML Engineer, Generative Video; Software Engineer, Agents; ML Engineer, Agentic Systems — and you're looking at a discipline that didn't have a name three years ago.

Traditional AI hiring splits cleanly into research scientists, ML engineers, and applied AI product managers. Mirage's open roles cut across all three. The ML Engineer, Generative Video position demands fluency in diffusion models and video generation architectures, but also requires building the infrastructure that serves those models at scale and the product interfaces that let non-technical creators direct them. That's not a research role with engineering support. It's a single job that expects you to own the stack from training data to user-facing output.

This hybrid profile shows up across the broader market. Netflix's INK division posted a Head of Technology role for its GenAI-native animation studio with a compensation range of $548,000–$811,250. The job description reads like three roles stitched together: pipeline architect, creative-tools product lead, and infrastructure strategist who also understands LoRA-based fine-tuning and data governance. Promise Studios in Los Angeles is hiring a VFX Supervisor, AI & Animation at $140,000–$165,000 who must bridge traditional photoreal VFX supervision with ML-driven asset generation and Unreal Engine real-time workflows.

These aren't prompt engineers. The prompt engineering career path, which upSkill's 2026 salary guide pegs at six figures for entry-level and north of $180,000 for senior specialists, is its own track, focused on optimizing inputs to existing models. Generative-media engineers build the models, the pipelines around them, and the products on top of both. The skill stack includes PyTorch or TensorFlow, cloud deployment on AWS or GCP, video-specific architectures like diffusion and consistency models, and enough editorial or animation literacy to know whether the output actually works for a professional creator.

That last piece is what separates this from standard ML engineering. A backend engineer optimizing a recommendation model doesn't need to understand shot composition. A generative-media engineer at Mirage does, because the product is the model's output and the output is a video a human intended to make. The feedback loop between infrastructure decisions and creative quality is direct and unforgiving.

The talent pool reflects this novelty. Most candidates aren't coming straight from ML PhD programs. They're migrating from adjacent fields — VFX pipeline engineering, creative-tool development, full-stack roles at companies like Runway or Pika — and picking up the missing pieces on the job. Studios like Toon2Tango in Munich and TrueShort in Los Angeles are explicitly hiring for this crossover, listing requirements that pair professional animation backgrounds with hands-on AI video tool experience.

Union Square as NYC's Post-Production AI Cluster

Mirage's headquarters in Union Square puts the company at the center of a neighborhood that has quietly assembled a concentration of generative-media and AI-video talent in New York City. The blocks around Union Square and the surrounding Flatiron district have become a default address for AI companies building at the intersection of creative tools and foundation models.

The neighborhood's draw is structural, not accidental. Union Square sits within commuting distance of both Midtown's enterprise clients and the creative agencies lining Chelsea and the Meatpacking District. For a company like Mirage, whose product has to serve both professional editors and first-time creators, that proximity matters. The talent pool includes engineers who've shipped media-production software, researchers who understand diffusion and transformer architectures for video, and product designers who know the difference between a timeline interface and a prompt-based workflow.

Mirage isn't the only AI-video player with a New York footprint. Runway lists an executive assistant role based in New York, though its engineering roles are remote. The broader NYC AI ecosystem, however, is far larger than any single company. Y Combinator's startup directory lists 111 AI companies headquartered in New York as of June 2026, with 16 tagged specifically as generative AI startups. Many of these are small (three to fifteen employees) but they span adjacent domains: Yarn helps sales and marketing teams make product videos with AI, Dream3D builds generative models for 3D scene simulation, and No Logo uses generative AI to match product designs with overseas manufacturers.

The city's institutional infrastructure is also moving. NYC Talent and CUNY announced a $5.3 million investment this year to launch CUNY Tech Futures, a program that will train more than 2,500 students per year across 12 campuses in AI-relevant skills. That pipeline is designed to feed exactly the kind of hybrid roles Mirage is hiring for — positions that require both ML chops and product intuition.

What makes Union Square distinct from, say, San Francisco's Mission district or Boston's Seaport is the vertical concentration. NYC's AI startups cluster around the city's incumbent industries: finance, media, healthcare, and law. Mirage's location places it in the media cluster, within walking distance of post-production houses, advertising networks, and the creative departments of companies like BuzzFeed, Vimeo, and the New York offices of every major agency holding company. When the company needs an ML engineer who understands why a 24fps output matters, or a product manager who's sat through a client review session, the local talent pool is the recruiting pitch.

The risk is competition for that same pool. Hebbia, AlphaSense, EliseAI, and Dataminr are all NYC-based AI companies hiring ML engineers and researchers at the same seniority levels. Mirage's $75 million round gives it runway to pay competitively, but in a market where Built In NYC lists hundreds of open AI roles across the city, compensation alone won't close candidates. The bet is that engineers who want to work on generative video will find a foundation-model lab more compelling than another fintech automation platform.

Whether Union Square becomes the recognized hub for generative-media engineering or remains one of several NYC clusters depends on whether Mirage and its neighbors can retain the talent they hire. The neighborhood has the companies, the capital, and the institutional support. The next twelve months of hiring data will show if it has the gravity.

Competitive Landscape: Who Else Is Hiring for AI Video Talent

The generative-video space is crowded with well-funded competitors, but the hiring signals tell a more nuanced story than the fundraising headlines. Zero G Talent's board data shows Mirage added 5 roles in the past 7 days. Runway added 0. OpenAI added 38, though OpenAI's roles span the full breadth of its operations, not just video.

The AI video generation tool landscape itself is in flux. OpenAI announced on March 24, 2026 that it is discontinuing the Sora app and API due to declining user interest, removing one of the highest-profile platforms from the market. Runway's Gen-4.5 has emerged as the professional's choice for production workflows, offering 4K output, character consistency, and a stable API. Pika 2.5 competes on speed and cost for short-form social content. Kling AI from Kuaishou offers up to 2-minute videos at low cost, though its Chinese hosting raises data considerations for enterprise users. Google's Veo 3.1 is emerging as a serious competitor with 60-second clips and strong motion consistency.

For talent, the implication is that the field is consolidating around a smaller number of production-ready platforms. Engineers building generative-video expertise are increasingly choosing between companies that have demonstrated unit economics and product-market fit (like Mirage with its $28.4 million in-app revenue) and those still searching for both. The $75 million round, paired with an existing 20-million-user base, puts Mirage in the former category. That distinction matters for recruiting: candidates with options tend to favor companies where the product already works and the hiring is about scaling, not speculating.

Working in AI? Zero G Talent tracks the openings: browse AI jobs, openings at OpenAI, Runway and Mirage, and the people building the field.