A Polymath In Progress: Daksh Miglani's Decade Of Building What Comes Next

Daksh Miglani started programming at 13. While still in school, he was already working on production software. More than a decade later, his career has taken him through enterprise systems, applied artificial intelligence, smart contracts, open finance, clinical technology and infrastructure for AI agents.

Those fields may look unrelated. Miglani's way of entering them is not. He tends to spot a new category early, approach it from several sides and stay long enough to build something that must work in the real world. His range is deliberate. It is how he has built his career.

"I have never been interested in building things that are easy to demo but fall apart in production," Miglani says. "The work I care about is the part that has to keep running when something important depends on it."

Starting early

Miglani entered the startup world while he was still in school. His early work included product engineering at StayUncle and later co-founding Adler, which built e-voting and e-auction software for clients including banks and audit firms.

The work gave him an early education in software with little room for error. A consumer app can recover from an awkward release. Voting and auction systems face a harder test. They must be secure, auditable and dependable when money or institutional trust is at stake.

Adler also showed Miglani that the engineering behind a product interested him as much as the product itself. That instinct would follow him into every field he entered next.

Applying AI before the current boom

Miglani moved into artificial intelligence before generative AI became a familiar part of startup vocabulary. He co-founded Nasch, an employee wellness platform that combined machine learning with psychology to help companies understand and support their employees.

Nasch used models including GPT-2 and BERT, placing the company in applied AI several years before the current wave of generative products. The work also brought Miglani recognition outside the startup community. He was selected for Youth Co:Lab India 2020, a programme associated with the United Nations Development Programme, Atal Innovation Mission and Citi Foundation.

The experience taught him how quickly a promising model can run into the messiness of real users, ambiguous language and organisational constraints. It also sharpened a preference that would shape his later work: learn a new technology by building with it before the category feels settled.

Designing his own education in Smart Contracts

Miglani saw decentralised finance as the next major technology cycle. He expected several years of intense experimentation, followed by enough institutional adoption for the technology to reach broader use.

Rather than study the field from one position, he joined Socket and InstaDapp at the same time and worked staggered shifts across the two startups. The schedule was demanding by design. It let him study different parts of the DeFi ecosystem in parallel and compress the learning curve. Both companies later grew into major Web3 protocols.

"I believed the next five years in DeFi would be extraordinary," Miglani says. "I expected a period of mania and experimentation, followed by the institutional adoption needed to make the technology viable."

He soon applied that education by founding a research lab, where he served as chief technology officer. He assembled the engineering team, led product development, raised capital and helped secure grants. The lab launched several products that reached production scale and attracted a community of users.

The quality of that work led to further contributions across several DeFi protocols. It also brought invitations to judge projects and support technical due diligence in the sector. The period gave Miglani experience with smart contracts, building a talented team, fundraising and product design in a market where a software error can carry an immediate financial cost.

Building for other builders

As Miglani built companies, he also became more involved with the people starting them. He mentored participants at Google Code-in, judged HackJNU and later served as a judge at the Bay Builders Hackathon and other developer events.

He was also accepted into Entrepreneur First's Bangalore programme. Entrepreneur First has used a talent-investing model since 2014 to help individuals form and build companies. Its backers include Stripe founders Patrick and John Collison, LinkedIn co-founder Reid Hoffman and Google DeepMind co-founder Demis Hassabis.

Miglani's work with young founders later grew into advising and angel support. He began working with Induced AI early, first advising its founders and later supporting the company as an angel. The browser automation startup went on to raise a $2.3 million seed round in 2023 from backers including Sam Altman, Peak XV and others.

He describes the aim simply: become the kind of support he lacked when he was starting out.

The shift did not pull him away from building. It gave him another angle on it. Working with founders exposed him to unfamiliar problems and forced him to explain technical and product choices clearly, often before a company had enough data to make the answer obvious.

Giving AI agents a way to act

By late 2024, Miglani had turned to a problem at the meeting point of AI and the open internet. Large language models could reason, write and interpret instructions, but an agent that needed to complete a task still had to operate browsers, handle authentication and survive changes in the software it used.

Miglani founded Session to work on that execution problem. The company built a proxy SDK and infrastructure for browser and device-based workflows. Its app partnership programme reached apps with over 10 million monthly active users.

The work addressed a gap between an impressive model response and a completed task. An agent can decide what to do, but it still needs dependable access to applications, data and authenticated sessions. Session focused on making that connection work beyond a controlled demonstration.

Going hands-on in clinical AI

Miglani's work with Neev made another part of his approach clear. He joined the clinical technology company as an advisor, then moved deep into the product and modelling work because the problem demanded it.

Neev built a clinical note-generation product for South Asian healthcare settings. Doctor-patient conversations in India can move across languages and dialects within a single sentence. A doctor, patient and attendant may each speak a different language. Medication names such as Trastuzumab, Adalimumab and Bevacizumab add another source of error. The audio may include coughing, sneezing, interruptions or a patient struggling to speak.

Miglani built a pipeline around those conditions. He says it outperformed existing transcription models on doctor-patient conversations. The work required more than adapting a general Western transcription system. The team built around the language, noise and clinical vocabulary of the environment in which the product would be used.

The resulting pipeline was used by clinics and hospitals in Bengaluru. For Miglani, the project combined a polymath's range with the depth of hands-on technical work. He entered as an advisor and ended up helping solve the modelling problems at the centre of the product.

A new role at Composio

Miglani will be joining Composio to lead the company’s work on AI agents and research.

Composio builds tools that let AI agents interact with external applications and services. Its platform handles integrations, authentication and application-specific behaviour that developers would otherwise have to build and maintain themselves.

The company raised $25 million in a Series A round led by Lightspeed Venture Partners in July 2025, bringing its total funding to $29 million. At the time, Composio said that more than 100,000 developers used its platform and that it served more than 200 companies, including Glean.

As AI agents move from producing answers to performing work, these connections determine what they can accomplish. The model may choose an action. The surrounding system has to authenticate it, send the right data and deal with whatever the external application does next.

Range as a method

No single technology defines Miglani's career. The consistency lies in how he approaches each one. He enters emerging fields early, studies them from several angles and goes deep enough to build products that have to work beyond a demonstration.

That method has taken him through enterprise software, applied AI, decentralised finance and healthcare. Each move has widened his range while sharpening the same instinct. Find the difficult technical problem before the category becomes obvious, then learn by building through it.

His career has never followed one lane, and that is the point. More than a decade after he began writing code as a teenager, Miglani is still following the same impulse: get to the next field early, understand it properly and make something real.