Software Engineer (Frontend)

Velt’s platform is trusted by industry leaders including Altana and Freshworks. We build the review and approval infrastructure for AI agents.

Here is the problem. Companies are now shipping agents that do real work. The agents write code, move money, email customers, change records. Almost nobody trusts them to do any of it unattended. So every one of these companies ends up building the same thing: a queue where a human looks at what the agent wants to do and says yes or no, plus the audit trail to prove who said yes. It takes them a quarter. It is not their product. We give them an SDK and it takes them an afternoon.

The catch is that our UI shows up inside their product, so it has to look like they built it. A dense enterprise console and a stripped down consumer app should both feel like the review queue was theirs all along. Today that means we expose a lot of knobs and hope we guessed right about which ones people want. That does not scale. There are infinite ways an app can look and we cannot enumerate them in advance.

So we want to attack it from the other side. Instead of a developer reading our docs and hand writing theme configs, an agent should read their design system, look at their code, generate the tokens and component overrides, show them what it will look like, and open a PR. Building that is the job. You would own both halves: the customization surface itself, and the agents that drive it.

This is an unusual frontend job, in a good way. Most frontend work is making one app look right. This is making our components look right inside apps you have never seen, written by people you have never met, and doing it automatically.

What you need to succeed:

• You have written a lot of CSS and are not afraid of it. Theming, design tokens, shadow DOM, style isolation, layout.
• You have built something real with LLMs. Prompting, tool calling, structured output, evals.
• Taste. You can look at a component and tell whether it is merely correct or actually right. This part is hard to teach.
• Strong communication skills, both written and verbal.
• You have shipped at least one library or SDK used by other developers.
• Deep understanding of modern web languages, API and frameworks.
• Relevant experience of atleast 1 year building two of these:
  • SDKs, widgets or component libraries that other developers use
  • scalable component libraries and design systems in JavaScript or TypeScript
  • headless or composable UI primitives with slots, render props or wrappers
  • AI tools that generate code or UI

What you will do

• Make every part of the Velt SDK UI customizable. Themes, design tokens, dark mode, layout overrides, slots, and fully headless pieces for teams that want to build their own UI on top of our logic.
• Build the agents that do the customization for people. They read a customer's design system and app, generate the theme and the overrides, preview the result and ship it. You design the tools, schemas, evals and guardrails that keep the output sane.
• Own the customization API. You decide what the props, tokens and escape hatches are, and how far someone can go before they have to write it themselves.
• Build the review and approval UI itself. Queues, diffs, audit trails, permissions, and the moments where a human steps in.
• Make sure our styles never leak into the host app and the host app never breaks ours.
• Write the docs, examples and playgrounds. If customization is not obvious in the first five minutes, we have failed.
• Build beautiful experiences and features for Velt that scale to millions of users.

Your first 30 days

• You will ship the initial agent-driven CLI that generates a Velt theme from a URL. Not a prototype: in production, in customers' hands, with a real customer themed end to end by day 30. We build with AI coding agents here, so we expect a working version inside week one. If a 30 day target for shipping a real product surface sounds unreasonable to you, this is not the right role.

We are looking for the following four fits:

1. Startup stage fit:

• You have experience working independently with abstract problem statements.
• You have experience thriving in unstructured and ambiguous environment.
• You get excited about just delivering, whether it’s a juicy challenging problem or a dry boring problem.
• You have the ambition to be the best in your craft.

2. Problem fit:

• You are excited about the problem we are working on and have an opinion about the problem/solution.
• You have a view on how humans should stay in control of what agents do.
• You have a view on where agents should be trusted to generate UI and where they should not.
• You care about developer experience and think of the SDK API itself as a product.

3. Experience fit:

• You have shipped at least one library or SDK used by other developers.
• You have worked at atleast one company for more than 2 years.
• You have deep understanding of modern web languages, API and frameworks.
• You have prior relevant experience of atleast 1 year building one of these:
  • SDKs, widgets or component libraries that other developers use.
  • scalable component libraries and design systems in JavaScript or TypeScript.
  • headless or composable UI primitives with slots, render props or wrappers.
  • AI tools that generate code or UI.
  • real-time web apps.
  • web apps for creating, drawing or editing.

4. Culture fit:

• You prioritize speed over perfection.
• You can explain difficult concepts to a 5 yr old.
• You are kind and can find common ground in difficult scenarios.
• You are hungry to learn and grow extremely fast.

Why you should join us

We are looking for someone who can hustle and is excited to go all in. This is not a 9-5 job. It's fast paced and demanding. We are working on very high impact products. This requires high velocity, low ego and lots of hunger. At Velt:

• You will be one of the first 10 employees and shape our engineering practices, products, company culture and business.
• You will get equity and participate in success of the company.
• You will work with an outstanding team.
• You will learn and grow at an accelerated pace.
• You will create your best work and shape how people stay in control of AI agents.
• Your code will get shipped to end users as soon as daily.

• Your work will matter. You will have a direct large impact on the company and our customers.

Submit the following when you apply:

• Why are you looking for a change right now?
• What excites you about the problem space and what is your point of view on that?
• Why are you the right fit?
• A link to a component library, design system or embeddable UI you built, and what you would change about its customization API today.
• Anything you have built with LLMs or agents, especially if it generated code or UI.

Intro call

Here we determine if the skills, experience, and motivations are aligned.

Round 1: Interview with the founder (60 mins)

In this interview, I will deep dive into how you think, solve problems & work with other people. We will go deeper into specific examples from your past work experiences.

Round 2: Technical interview (60 mins)

You will get a technical problem to work on.

Round 3: One take home project (Few hours)

We will provide you with a small project to work on. You can work on this part time at your own pace. During the project you will have access to us, you can ping us if you have any questions.

Final Round: Present your project & Offer (45 mins + 45 mins)

In this round you present your project to us. Here, we are mostly interested in the approach you took to solve the problems you faced, any deep insights you learnt etc.

Within 24 hrs of this round, we will make a decision and update you.

If we make an offer, we will discuss the compensation, equity and benefits at length with you on a call. Our offer conversations are longer than most other companies. That's because if you've made it this far, we're excited about working together, and we want to spend as long as it takes on anything you'd like to discuss.