Softlight
Infrastructure for AI coding agents to produce interfaces consistent with a company's design system.
Website: https://softlight.com/
Cover Block
From the public record
| Field | Value |
|---|---|
| Name | Softlight |
| Tagline | Infrastructure for AI coding agents to produce interfaces consistent with a company's design system [softlight.com] |
| Headquarters | New York City, United States |
| Stage | Seed |
| Business model | SaaS |
| Industry | Other |
| Technology | AI / Machine Learning |
| Geography | North America |
| Growth profile | Venture Scale |
| Founding team | Co-Founders (3+): Rohan Kapoor, Aditya Raju, Ashwin Madavan [Rohan Kapoor, October 2025] |
| Funding label | Seed |
| Total disclosed | ~$3,200,000 [Rohan Kapoor, October 2025] |
Links
From the public record
- Website: https://softlight.com/
- LinkedIn: https://www.linkedin.com/jobs/view/you-create-a-role-in-new-york-ny-at-softlight-4336422224
- Other: https://jobs.ashbyhq.com/softlight
The Short Version
PUBLIC Softlight is building software infrastructure that aims to make AI coding agents generate interfaces that match a customer’s design system and product conventions, a narrow but timely problem as code generation tools move from prototyping into production workflows [softlight.com] [LinkedIn, October 2025]. The company merits attention now because it has already positioned around a specific bottleneck in enterprise AI adoption, raised $3.2 million in disclosed seed funding, and is hiring across core engineering roles in New York, all signs of an early but active buildout [Rohan Kapoor, October 2025] [Ashby] [LinkedIn, October 2025].
The public record suggests some product evolution rather than a static pitch: earlier recruiting and founder materials described Softlight as building product and design agents that help teams discover what to build, learn a company’s business context, and generate design alternatives, while the current website centers more explicitly on infrastructure for coding agents [LinkedIn, October 2025] [softlight.com]. That shift is worth tracking because it may indicate either sharpening market focus or broader experimentation around the same underlying asset, namely company-specific product and design context [softlight.com] [LinkedIn, October 2025].
Softlight’s founding team includes Rohan Kapoor, Aditya Raju, and Ashwin Madavan, with public materials linking the team to prior work at Meta, Uber, Citadel, Faire, Runway, and Two Sigma, although most of those background details are founder- or profile-sourced rather than independently reported [RocketReach] [Aditya Raju - Founder @ Softlight, 2026] [Ashwin Madavan - CTO @ Softlight, 2026] [LinkedIn, October 2025]. That pedigree is relevant because the company is asking customers to trust it with design-system logic and internal product conventions, which tends to reward technical credibility and product judgment early [softlight.com] [LinkedIn, October 2025].
On the commercial side, Softlight presents as a SaaS company and says access is currently limited to a closed group of high-growth companies while it scales compute, with founder materials also referring to design partners ranging from growth-stage startups to Fortune 500 companies [softlight.com] [Rohan Kapoor, October 2025]. The financing record in the available sources is still thin, but the October 2025 seed announcement gives Softlight enough capital to support a small founding team and an initial go-to-market motion without yet answering who led the round or how concentrated customer demand is [Rohan Kapoor, October 2025].
Over the next 12 to 18 months, the key public signals to watch are straightforward: whether Softlight names customers or case studies, whether its positioning settles around developer infrastructure versus upstream product ideation, and whether hiring converts into a repeatable product surface rather than bespoke design-partner work [softlight.com] [Ashby] [LinkedIn, October 2025]. If those signals firm up, the company could occupy a useful control layer between general-purpose coding agents and the design consistency requirements of larger software teams; if they do not, the main question will be whether the problem is acute enough to support durable software spend at scale [softlight.com] [LinkedIn, October 2025].
Unconfirmed -- This section relies primarily on company website, founder posts, recruiting materials, and profile pages, with no independent named-publisher reporting located in the provided sources.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | SaaS |
| Industry / Vertical | Other |
| Technology Type | AI / Machine Learning |
| Geography | North America |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+) |
| Funding | Seed, total disclosed about $3.2 million [Rohan Kapoor, October 2025] |
The Company in Brief
PUBLIC
Softlight presents as a New York-based software startup focused on a narrow but increasingly relevant layer of the AI developer stack: getting machine-generated interfaces to match a company’s own design language rather than default patterns [softlight.com]. The current website describes the company as building infrastructure that makes coding agents "design like your company," and frames access as limited to a closed group of high-growth companies while it scales capacity [softlight.com].
The public record is still thin on formal company history. The available company materials identify New York City as headquarters and name Rohan Kapoor, Aditya Raju, and Ashwin Madavan as the founding team, but do not provide a founding date or legal entity details in the cited sources for this section [softlight.com].
The clearest dated milestone in the public record is October 2025, when founder-linked materials announced a $3.2 million seed financing and described an earlier product framing around product and design agents that help teams discover what to build [softlight.com]. Since then, the website positioning appears more specific: infrastructure for coding agents to generate interfaces consistent with a customer’s design system and product conventions [softlight.com]. That shift does not by itself establish a full pivot, but it does suggest message tightening toward developer workflow and interface consistency.
Unconfirmed -- This section relies primarily on company website materials, with no Crunchbase or state filing corroboration used here.
What They Have Built
MIXED
Softlight’s public product description is narrow, but the core use case is clear. The company says it is building infrastructure that makes coding agents “design like your company,” with the stated goal of helping AI coding tools generate interfaces that match a customer’s existing design system and product conventions [softlight.com]. That framing matters because it places Softlight less in generic code generation and more in the consistency layer between AI output and a company’s established product language [softlight.com].
The earlier public materials point to a somewhat broader product surface. In October 2025 hiring posts and founder announcements, Softlight described itself as building product and design agents that help teams discover what to build, learn a company’s business context, generate product ideas, and produce design alternatives for product managers, engineers, and designers [LinkedIn, October 2025] [Rohan Kapoor, October 2025] [Ashwin Madavan, October 2025]. Read conservatively, the positioning appears to have evolved from product discovery and design assistance toward infrastructure for coding agents, rather than away from design entirely [softlight.com] [LinkedIn, October 2025]. The company also says access is limited to a closed group of high-growth companies while it scales compute, which suggests the product remains in a constrained deployment phase rather than broad self-serve availability [softlight.com].
The technical stack is only partially visible from public evidence, so any deeper read has to stay modest. The product is plainly AI-native from the company’s own descriptions, and the open roles for Founding AI Engineer, Backend Engineer, Frontend Engineer, and Product Engineer imply an architecture spanning model-facing systems, application infrastructure, and user-facing product workflows (inferred from job postings) [Ashby] [LinkedIn, October 2025]. What the public record does not yet establish is just as important: there is no verified public demo, no disclosed model strategy, no named integrations, and no technical documentation on how Softlight ingests design systems or enforces UI consistency across coding agents in production [softlight.com] [Ashby].
Unconfirmed -- This section relies primarily on company website copy, founder posts, and recruiting materials, with no independent technical documentation or third-party product verification.
Market Size and Demand
PUBLIC
The market matters now because AI-assisted software creation is moving from code generation toward production-ready product assembly, and that raises a narrower but commercially meaningful problem: how to make generated interfaces conform to a company’s design system, product conventions, and internal standards [softlight.com] [LinkedIn, October 2025].
Public evidence on Softlight’s precise addressable market is thin, so the cleaner approach is to bracket it through adjacent categories rather than force an unsupported TAM. The company’s current positioning sits between AI coding tools and enterprise design-system software, with earlier materials framing the product as a product-and-design agent for idea generation and interface exploration [softlight.com] [LinkedIn, October 2025]. That places Softlight in an analogous market shaped by spending on developer tooling, product design software, and workflow automation for product teams, but no named third-party market-sizing report was captured in the supplied sources for any of those segments.
Demand drivers are easier to see than market size. Softlight’s own materials point to the friction between generic AI output and company-specific product language, especially where teams want AI coding tools to generate interfaces that reflect existing design systems rather than produce generic layouts [softlight.com]. The hiring and founder posts also suggest a broader workflow ambition, where product managers, engineers, and designers use agents to discover ideas, generate alternatives, and accelerate handoff across functions [LinkedIn, October 2025]. If that problem is real inside customers, the budget could come from several existing lines rather than a new category, including developer productivity, design tooling, and product operations.
The adjacent markets are material because they define both Softlight’s entry points and its substitution risk. On one side are AI coding assistants and agentic developer tools, which already promise faster software production; on the other are design platforms and internal design-system workflows that govern consistency, review, and reuse. Softlight appears to be aiming at the layer between them, where business context and design constraints are translated into interfaces a coding agent can use [softlight.com] [LinkedIn, October 2025]. That is a plausible wedge, but it also means the company may be pulled into categories with incumbents that own the developer seat, the designer seat, or both.
Macro conditions are supportive in a narrow sense. The company says access remains limited to a closed group of high-growth companies while it scales compute, which implies the product depends on model availability and inference economics as much as user demand [softlight.com]. At the same time, the absence of named customers or deployments in the public record means there is not yet enough evidence to say whether this is a broad enterprise budget line or still a design-partner use case [softlight.com] [LinkedIn, October 2025]. Regulatory exposure appears indirect rather than category-defining based on the sources provided: the more immediate commercial constraint is whether companies are willing to feed internal design systems, product conventions, and business context into external AI workflows.
| Market lens | What the public record supports | Source basis |
|---|---|---|
| Core category | AI infrastructure for coding agents to design like a customer’s company | [softlight.com] |
| Functional buyer surface | Product managers, engineers, and designers | [LinkedIn, October 2025] |
| Primary job to be done | Generate interfaces consistent with an existing design system and product conventions | [softlight.com] |
| Adjacent spend pools | Developer tooling, design tooling, product workflow software | Inference from product positioning, based on [softlight.com] [LinkedIn, October 2025] |
| Current commercialization posture | Closed group of high-growth companies while compute scales | [softlight.com] |
The table shows why the market case is interesting but still early. The product maps onto existing software budgets and a visible workflow bottleneck, yet the public evidence does not establish segment size, purchasing motion, or whether this will settle as a standalone category versus a feature inside broader AI development platforms.
Inferred, not confirmed -- This section relies mainly on company website and recruiting or founder materials, with adjacent-market interpretation inferred from product positioning rather than supported by independent third-party market reports.
Who Else Is Fighting for This
MIXED Softlight is positioning itself less as a standalone design tool and more as infrastructure that sits underneath AI coding agents, a narrower lane than broad product-design copilots but one that could matter if interface generation becomes a standard step in software creation [softlight.com] [LinkedIn, October 2025].
The public record is thin on named direct competitors, so the competitive map has to be drawn by category rather than by a clean company list. At one end are incumbent design and product workflow systems that already own designer and PM attention, even if the sources here do not name them directly. In the middle are challengers building AI-native product or design agents, which is the lane Softlight itself described earlier when it said it was building agents that help teams discover what to build for PMs, engineers, and designers [LinkedIn, October 2025]. Adjacent substitutes are the general AI coding tools that can already generate interfaces, but often without a company-specific design system or product conventions, which is the gap Softlight says it is addressing today [softlight.com].
Softlight's edge, on the evidence available, appears to be specificity rather than breadth. The company is not claiming to replace the whole product stack; it is claiming that coding agents need a design layer tuned to a customer's existing interface patterns and business context [softlight.com] [LinkedIn, October 2025]. That wedge could become durable if Softlight accumulates proprietary customer context, design-system mappings, and feedback loops that make its outputs visibly better over time. It is also a perishable edge, because the same value proposition could be pulled into larger coding-agent platforms if they decide design-system compliance is a core feature rather than an extension point.
The exposure is straightforward. Softlight does not appear, from public materials, to own the primary developer channel, the primary design channel, or a named enterprise distribution partner [softlight.com] [Ashby]. If general-purpose coding assistants build native support for company-specific UI conventions, Softlight could face pressure from the platform layer above it. The company is also exposed to category drift from its own earlier positioning: product discovery, idea generation, and design alternatives are broader promises than interface-consistency infrastructure, and broader categories tend to attract more well-capitalized entrants over time [LinkedIn, October 2025].
The most plausible 18-month scenario is a split market. A winner if company-specific context becomes the gating factor for AI-generated software is Softlight itself, because its entire public narrative is built around making generated interfaces look and behave like the customer's product rather than a generic template [softlight.com]. A loser if that does not happen is Softlight as well, because the sources do not yet show a public moat in distribution, customer logos, or independent ecosystem validation that would protect it if generic coding platforms close the quality gap quickly [Rohan Kapoor, October 2025] [softlight.com]. On current evidence, the company's competitive promise is real enough to watch, but it still rests more on product direction than on publicly verified market control.
Unconfirmed -- This section relies primarily on company website language, founder LinkedIn posts, and recruiting materials; no named competitors or independent market-share benchmarks were confirmed in the provided public sources.
Opportunity
PUBLIC
The prize here is large if Softlight can become the infrastructure layer that makes AI-generated product interfaces production-acceptable inside real companies, because that would place it on the path of every team trying to turn code generation into shipped software rather than disposable prototypes [softlight.com] [LinkedIn, October 2025].
The headline opportunity is narrower, and therefore more credible, than a generic claim to "AI for product design." Softlight's current positioning is explicitly about making coding agents design like the customer's company, with outputs aligned to existing design systems and product conventions [softlight.com]. That matters because the practical bottleneck for AI coding tools is not only code generation, it is whether the output matches how a company already builds and ships interfaces. Softlight's earlier hiring language also suggests the company has been working at the product and design layer for some time, first around idea generation and design alternatives for PMs, engineers, and designers, and now around infrastructure for coding agents, which reads less like a category jump than a wedge tightening around a harder enterprise need [LinkedIn, October 2025].
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Design-system control plane | Softlight becomes the default layer enterprises use to steer AI coding tools toward approved UI patterns, components, and product conventions | A successful rollout with current design partners pushes the product from closed access to broader commercial availability | The website positions the product as infrastructure for coding agents, and the company says access is already live with a closed group of high-growth companies, which implies active testing rather than concept-stage positioning [softlight.com] |
| Product-to-code workflow platform | The company extends from interface consistency into upstream product discovery and design generation, owning more of the workflow between idea and shipped UI | The earlier product-discovery agent workflow becomes a commercial module rather than a discarded concept | Founder and hiring materials described Softlight as learning a company's business, generating ideas, and designing alternatives, which suggests adjacent product surfaces already exist in the team's thinking [LinkedIn, October 2025] |
| Enterprise design partner expansion | A handful of recognizable large-company deployments turn Softlight into a standard purchase for design-led software teams adopting AI-assisted development | A named Fortune 500 customer or product executive reference de-risks procurement and drives peer adoption | The company has publicly said its design partners include growth-stage startups and Fortune 500 companies, even though those customers are not yet named [LinkedIn, October 2025] |
The compounding logic is straightforward if the product works as described. Each customer gives Softlight another environment in which to learn how real companies encode business context, design systems, and product conventions into software outputs [softlight.com] [LinkedIn, October 2025]. If that produces better alignment between generated code and shipped product, Softlight could move from being a useful layer for experimentation to a workflow dependency for teams that want AI-generated interfaces to pass internal review faster. The early signs are thin but directionally relevant: the company is recruiting across AI, backend, frontend, and product engineering in New York, which is consistent with building both model-adjacent infrastructure and developer-facing product surfaces rather than a single-feature tool [Ashby] [LinkedIn, October 2025].
The size of the win is hard to pin down from public evidence because no verified market-sizing study or direct public comparable appears in the source set. Even so, the upside case is not abstract. If Softlight becomes a standard layer between AI coding tools and enterprise product teams, the company could plausibly resemble a workflow-infrastructure software asset with strategic value to developer-platform vendors, design-tool incumbents, or model-platform companies that need better enterprise-grade interface generation. With only the public record available, the cleanest way to state it is qualitative rather than numeric: this is a scenario in which Softlight could matter far beyond a seed-stage tool, because it sits at the control point between generated code and brand-consistent software output [softlight.com] [LinkedIn, October 2025]. That is a scenario, not a forecast.
Unconfirmed -- This section relies primarily on company website statements, founder posts, and recruiting materials, with limited independent corroboration.
Sources
From the public record
[softlight.com] Softlight | https://softlight.com/
[Rohan Kapoor, October 2025] Product design is one of the most misunderstood fields in tech. It’s... | https://www.linkedin.com/posts/rohanneilkapoor_product-design-is-one-of-the-most-misunderstood-activity-7388972842429845504-ZnoH
[LinkedIn, October 2025] Founding AI Engineer (NYC) at Softlight | https://www.linkedin.com/jobs/view/founding-ai-engineer-nyc-at-softlight-4319487550
[Ashby] Softlight jobs | https://jobs.ashbyhq.com/softlight
[RocketReach] Rohan Kapoor, Co-Founder and CEO at Softlight | https://rocketreach.co/
[Aditya Raju - Founder @ Softlight, 2026] Aditya Raju - Founder @ Softlight: AI that discovers what to ... | https://www.linkedin.com/in/aditya-raju-1a3132104/
[Ashwin Madavan - CTO @ Softlight, 2026] Ashwin Madavan - CTO @ Softlight: AI that discovers what ... | https://www.linkedin.com/in/ashwin-madavan/
[Ashwin Madavan, October 2025] We’re building AI agents that think like world class product... | https://www.linkedin.com/posts/ashwin-madavan_were-building-ai-agents-that-think-like-activity-7388973194847735808-PlaU
Articles about Softlight
- A Closed Group of High-Growth Companies Is Testing Softlight's Design Agents — The New York startup, founded by alumni of Meta, Uber, and Citadel, is betting that AI-generated software must learn a brand's visual language.