The Infra Play

The Pipeline India Optimized For Does Not Usually Produce Category Builders

Madhav Jha and Mukund Jha are twin brothers whose paths through technical education diverged early — and whose complementary trajectories converged, years later, into one of the AI coding market's most precisely positioned companies.

Madhav went to IIT BHU, where he developed the research-depth orientation that would later define Emergent's architectural thesis. He went on to earn a PhD from Penn State and completed a postdoctoral fellowship at Sandia National Labs — the kind of formation that trains a person to ask not just how a system works but whether the system is designed correctly in the first place. He was a founding member of the Amazon SageMaker research team before co-founding Emergent. His thinking is structural. He sees infrastructure problems before they surface in production.

Mukund took a different route. He completed his B.Tech at Motilal Nehru NIT, then earned a Master's from Columbia Engineering. He moved into the operator world and eventually became CTO at Dunzo — one of India's most demanding fast-growth environments, a company that ran delivery logistics at scale across Indian cities and broke infrastructure problems on a near-daily basis. He absorbed the chaos of systems failing under load. He knew what it cost to debug opaque infrastructure. He saw, in direct terms, what the absence of production clarity does to a team trying to move fast.

Together, the two formed a founding pair with a specific and rare combination: one who had spent years thinking about how systems should be built, and one who had spent years watching what happens when they are not. That pairing is not accidental. It is the reason Emergent's core thesis — production infrastructure as a first-class concern — felt, to the founders, less like a product decision and more like an obvious conclusion.

FOUNDERS Madhav Jha and Mukund Jha, twin brothers and co-founders of Emergent. Madhav: IIT BHU, PhD Penn State, Sandia National Labs postdoc, founding member of Amazon SageMaker research team. Mukund: MNNIT, MS Columbia Engineering, former CTO at Dunzo. Launched Emergent in August 2024.

A Decade of Tools That Reduced Friction Without Eliminating Dependency

To understand what Emergent was entering in August 2024, you need to understand what the decade before had produced. Bubble launched in 2012. Webflow launched in 2013. Both were genuine advances. They enabled non-technical founders, small-to-medium businesses, and solo operators to build landing pages, internal tools, and functioning web products without writing code. Over the following decade, approximately 100-plus micro-agencies built service businesses on top of these platforms globally. Tens of thousands of SMB sites were enabled. The combined industry valuation entered the billions.

But the category had a structural ceiling it could never break through. Bubble and Webflow reduced technical friction. They did not eliminate technical dependency. An SMB founder still needed templates, still needed to understand workflows, still needed to hire a semi-skilled operator who knew the platform well enough to implement what the business actually required. The tools democratized access to building, but they preserved a layer of intermediary that any truly self-sufficient business would want to remove.

What this meant in practice: an e-commerce founder building a custom internal tool in 2021 still needed to either learn Bubble deeply or pay an agency $3,000–$15,000 to build it for them. The tool existed. The knowledge to use it well did not distribute evenly. No-code reduced the cost and complexity of building. It did not fundamentally change who could build without external help.

November 30, 2022. The Market Changed. The Infrastructure Problem Did Not.

ChatGPT launched on November 30, 2022. The AI coding category that followed was swift, crowded, and largely convergent. Coding became one of AI's earliest and strongest domains because the feedback loop is exceptionally clear: code either runs or it does not. There is no ambiguity in the output. This clarity made it a natural proving ground for large language models, and the ecosystem that followed reflected that.

Anthropic deepened the research frontier. Perplexity AI reframed search. Replit, Lovable, and Bolt.new each staked a position in the AI-assisted coding market, offering various flavors of the same fundamental promise: describe what you want to build, and the AI will build it for you.

The gap was specific and consequential. The first generation of AI builders had no backend clarity — generated code that connected to databases via undocumented logic. No source transparency — outputs were black boxes, correct until they were not, with no legible path to debugging. No rational architecture enforcement — the AI would produce code that worked in demo but fragmented under load, under integration, under anything that approximated real-world conditions. Most of these tools were optimized for the moment of demonstration, not the months of operation that followed.

CONTEXT Madhav and Mukund Jha. Their combined background — research-oriented systems thinking from Madhav's academic and SageMaker formation, and operator-level infrastructure experience from Mukund's CTO tenure at Dunzo — positioned them to see the demo-versus-production gap that competitors were systematically ignoring.

This is the distinction Madhav and Mukund identified. Not that AI could not code — clearly it could. But that AI coding without production infrastructure was a different product category entirely from what the market actually needed. Demo machines are not infra companies. The market was full of demo machines.

August 2024. A Structural Thesis, Not a Feature List.

Emergent's first product launched in August 2024. The timing is worth noting: eighteen months after ChatGPT, long enough for the first generation of AI coding tools to have made their demo-versus-production gap visible, short enough to still be early in what would become a major category. The founders had watched the ecosystem develop. They had identified the ceiling. They built for what was on the other side of it.

PRODUCT Emergent's web-based interface combines a VS Code-style editor with AI-assisted scaffolding, full backend visibility, and native GitHub integration — enabling developers to ship production systems, not just working prototypes.

The thesis was direct: if AI cannot reason structurally about a system, it cannot produce a system that survives production. The product was built around what that thesis required. Full frontend code access — the output was legible, editable, owned. Full backend visibility — database logic was transparent, not abstracted away. A web-based VS Code interface — developers worked in an environment they already knew. Native GitHub push — the output moved directly into existing engineering workflows. Exportable infrastructure — the system could leave Emergent and run elsewhere. Architecture that could be edited — not locked to a generation pattern.

The decision to show all code — rather than generate a running app and hide the mechanics — was a deliberate counter-positioning against every other tool in the market. Competitors had implicitly argued that users did not need to see the code, that the abstraction was the value. Emergent's implicit argument was the opposite: the abstraction is the liability. Give users the code, and you give them control. Control reduces switching risk. Control builds trust. Trust reduces churn.

The founders were also direct about AI's current limitations. Rather than positioning the product as a complete replacement for engineering judgment, Emergent acknowledged that AI cannot yet do one hundred percent of the work at production quality. The response to that limitation was not to hide it — it was to design the product around it. AI generates the scaffold. The developer owns the system. That division of labor is honest, and honest products build different kinds of retention than products built on overstatement.

August 2024 to February 2026: Eighteen Months of Structural Growth

The Arc: August 2024 – February 2026

The table below maps each material period across five parallel tracks: revenue status, funding events, valuation, product releases, and organisational changes. Figures unavailable via public disclosure are marked accordingly.

What Each Funding Phase Unlocked

Capital events at Emergent did not produce marketing spend or headcount inflation. Each phase funded a specific product layer, which produced a specific revenue inflection. The sequence below maps that relationship mechanically.

As of February 2026

Four Mechanisms. One Coherent System.

Emergent did not have a traditional go-to-market motion. No SDR team. No cold outreach at scale. No conference booths. What it had was a set of interlocking mechanisms that each reinforced the others — and a founding team willing to be personally visible in a market where infrastructure companies rarely make their founders the face of the product.

Built in Bangalore. Monetized in San Francisco and London.

Emergent's geographic structure is not accidental. Seventy to eighty percent of the team sits in Bangalore. The United States and Europe account for approximately seventy percent of revenue. India is the fourth-largest revenue market and the fastest-growing region by user count. This configuration reflects a specific thesis about where AI infrastructure companies can operate most effectively — and most durably.

The India thesis Emergent represents is one that a growing number of category-defining companies are running. The arbitrage is structural: India produces world-class engineering talent at a cost structure that Western startups cannot replicate, while the primary monetization markets remain the US and Europe. This is not a temporary condition. It is a durable advantage for founders who understand how to operate across both geographies simultaneously.

India is also where the developer community is growing fastest. The country's technical education pipeline produces hundreds of thousands of software engineers annually, creating a natural acquisition surface for developer tools. Emergent's university GTM flywheel operates most efficiently here. The fastest-growing user segment and the most cost-efficient engineering team occupy the same geography. That is a compounding structural advantage.

Demo Builders Versus Infra Builders. The Distinction That Defines the Category.

The AI builder market has a division that is not always stated clearly. There are tools optimized for the moment of demonstration — producing working prototypes quickly, impressing a stakeholder, validating an idea. And there are tools optimized for the months that follow: deployment, iteration, debugging, integration with existing systems. These are different products serving different moments in the development lifecycle.

The comparison above is not a ranking of overall quality. Lovable, Bolt.new, and Replit have built real businesses serving real needs. The point is category distinction. A founder who needs a landing page quickly, or a non-technical entrepreneur validating an idea with a clickable prototype, is well-served by demo-optimized tools. A developer who needs to ship a production system — something deployed, maintained, debugged, and integrated — is not. Emergent built for the second person.

Five Lessons. Framework-Level. Operator-Actionable.

Sources: Emergent company communications and founder LinkedIn posts; SWE-Bench benchmark public leaderboard (swebench.com); India startup ecosystem reporting (The Ken, Inc42, Economic Times Tech); Raj Shamani podcast archives; public funding announcements where available. ARR and valuation figures are reported or estimated based on available founder and investor disclosures. User count figures from company-stated milestones. Competitive table based on publicly available product documentation as of Q1 2026. Team size and geographic breakdown estimated from LinkedIn data and public interviews. Founder background details sourced from LinkedIn profiles and public interviews. All figures marked "reported" or "estimated" where direct company disclosure is unavailable.