Smoot logo

Smoot

Software & ITApp development

Sobre nosotros

Because every brand and campaign goal is unique, so should it's targeting. Smoot's new AI-powered technology provides advertisers with fully personalized generative AI models for their campaigns, resulting in eliminating +30% in waste impressions compared to a traditional programmatic campaign.

With Smoot, our ambition is to redefine the paradigm of programmatic media buying. Through the use of artificial intelligence, we aim to perfectly match brand values with the consumer's emotional context, championing the cause of conscious marketing.

"Smoot," derived from "smooth," represents the seamless, effortless experience we aim to provide to our clients and their customers. But beyond the surface, it symbolizes our commitment to emotional smoothness, where ads are tailored to the viewer's emotional state, thereby crafting a more natural and engaging experience.

Datos y curiosidades

AI-First from Day One: Smoot was one of the first companies to apply generative AI to brand advertising and intelligent media buying. Built for Results: Every product, format, and insight is designed to deliver measurable impact. Human Expertise + Machine Intelligence: The perfect blend of strategy and technology.

¿Qué estamos buscando?

We value people who: - Take ownership: own problems and are accountable for results from day one. - Focus on impact: measure everything and prioritize what delivers results. - Communicate openly: speak clearly and directly, without unnecessary politics. - Keep learning: experiment fast, learn and double down on what works. - Automate by default: use AI to multiply their capabilities, not replace critical thinking. - Work as a team: value respect, empathy and collaboration

Ofertas de trabajo en Smoot

Smoot SPX is an ad exchange (SSP) for connected TV and in-app advertising. It handles billions of ad requests a day, runs a real-time auction for each one and reports every dollar to the people who run the business.

We are adding an engineer who builds with AI agents instead of typing every line. You take a problem from the tech lead, turn it into a clear spec and direct Claude Code to build it. Then you do the part that cannot be delegated: read the diff, test it, and prove with real traffic data that it does what it should.

You will ship features across the backend and the operations console, working directly with the founder / tech lead. It is not a machine-learning role and not a DevOps role.

What you'll work on

  1. Traffic quality. Tools that detect invalid or low-value traffic and stop it before it reaches a buyer.
  2. Integrations. Third-party measurement and verification services, partner reporting APIs and data feeds: bring their data in and reconcile it with ours.
  3. Dashboards and reports. New console screens for supply, demand and revenue, end to end: the query, the API route and the UI.
  4. Optimisation. Features that decide which traffic is worth sending to which buyer, and how much of it (pacing, throttling, caps).
  5. Tools for the operations team. Bulk actions, validation checks and alerts that remove manual work and mistakes.
  6. Automation. Scheduled reports, partner exports and data checks that run on their own.
  7. AI for the team. Reusable agent skills and playbooks, so an analysis that takes an hour takes minutes.
  8. Investigations. "Why did this number move?" answered with data, then turned into a change.

How you'll work

AI agents write most of the code. You answer for all of it.

  1. Understand before you change. Read the code path end to end (backend, data and UI) and trace what depends on it.
  2. Design first. Data model, API contract, failure modes and behaviour at 10× the traffic, before the agent writes a line.
  3. Orchestrate agents. Split the work into tasks an agent can finish, give each one the right context, and use independent agents to review the result.
  4. Verify with evidence. Automated tests, a local run against real databases in Docker, and queries on production data that show the change does what it should.
  5. Ship safely. Small pull requests, new behaviour behind a flag. The tech lead reviews and merges.
  6. Close the loop. After release, read the numbers and say whether it worked.

Tech stack

  1. AI tooling: Claude Code: agents, sub-agents, skills, project rules and memory. Level needed: Daily driver (Cursor or Codex experience transfers).
  2. Backend: Node.js, Fastify, JavaScript. Level needed: Read, review and debug.
  3. Frontend: React 18, TypeScript, Vite, Tailwind, shadcn/ui, TanStack Query, Recharts.Level needed: Read, review and debug.
  4. Data: SQL on PostgreSQL (configuration) and ClickHouse (reporting); Redis (live counters, block Level needed: Write and check your own queries lists); analysis scripts in JavaScript. or Python
  5. Testing: Node test runner, Vitest + Testing Library, Docker Compose for a local Postgres and Redis. Level needed: Run and extend
  6. Workflow: Git, GitHub pull requests, GitHub Actions. Level needed: Fluent.
  7. Platform: AWS (EC2, Kinesis, S3, CodeDeploy), Terraform. Level needed: Context only: you will not operate it.
  8. Ad tech: OpenRTB 2.x, VAST, CTV and in-app, invalid traffic (IVT), supply chain (schain, ads.txt). Level needed: Learned on the job; a plus if you bring it

What we expect from you:

Attitude

  1. You want to understand the business, not only the code. How an exchange makes money, what a buyer pays for, why a partner complains. You keep asking until you could explain it yourself.
  2. You own the outcome. The job is done when the number moved, not when the pull request merged.
  3. You go to the cause. When something breaks or a metric drops, you find what caused it instead of patching what you see.
  4. You don't assume. You check the data first, and you say "I don't know yet" when you don't.
  5. You start simple. The smallest change that solves the problem, then you improve it.
  6. You learn fast and like it. A new domain and new tools every month: for you that is the fun part.
  7. You speak up early. On a live exchange a wrong guess costs money, so when you are unsure you ask before you act.
  8. You communicate short. Conclusion first, then the evidence.

Experience

  1. 2–4 years shipping production web software with JavaScript or TypeScript and SQL.
  2. You can show something you shipped mostly through coding agents, and explain how you checked it.
  3. Good with data: you notice when a number cannot be right.
  4. You build for a non-technical operations team: the simplest screen that answers the question.
  5. Clear, short written English. Code, UI and pull requests are all in English.

Nice to have

  1. Ad tech experience (SSP, DSP or exchange).
  2. An analytics database (ClickHouse, BigQuery, Snowflake).
  3. Something built on an LLM API (Claude API, tool use, agents).
  4. Spanish.
HybridBarcelona, SpainClaude codeAI agentVitestClickHouseFastifyTypeScriptSQLReactNode.JSJavaScriptGithub
Ver oferta