Founding Data & AI Engineer, Creative Engine
Company hidden until unlock
- Work model
- Remote · LatAm-friendly
- Location
- Brazil (Remote)
- Posted
- 7d ago
pythondjangorestpostgresredisawsdockerllmtailwindci/cdgithub actions
Company hidden until unlock
🏔️ ABOUT SIERRA STUDIO Sierra Studio is an innovative product development firm dedicated to empowering founders and business leaders in their quest to create impactful, meaningful products. Our team of experienced professionals is committed to delivering high-quality solutions that drive business growth. We prioritize a culture of good vibes, open communication, and continuous learning, providing our team members with the opportunities and resources they need to thrive. This role is with one of the teams we help build through our staffing partnerships. When you apply, Sierra manages the hiring process, but you'll be hired directly by and work within the client company. 🏢 ABOUT THE COMPANY Our hiring partner is a performance creative agency that makes ads for brands on TikTok, Meta, and TikTok Shop. They work with 2,000+ creators and produce thousands of videos per year. Their internal portal handles the entire workflow: creator onboarding, contracts, campaign management, content review, and payments. They're now building a Creative Engine that turns creative strategy from gut instinct into a data-driven process. Data today lives in too many disconnected places — the Creative Engine brings it all into one central home, across performance creative, TikTok Shop, and TikTok LIVE, so the team can derive real insights instead of digging through separate platforms. AI watches and tags every video they produce, breaking each one into the hooks, formats, and patterns that drive results, and connects that to the sales and engagement data flowing out of Shop and Live. It all feeds one queryable system, a "brain" a strategist can ask "what are the top hooks for beauty this week?" and get a real answer back. The team reaches it through dashboards, Slack, and Cowork. It starts as the internal engine the creative team runs on, and over time clients log in to it too — this is the core of where the company is headed. How they work: - Small team, high ownership. You own your work end to end, from idea (or data model) to deployed UI. No committees. - AI-assisted everything. Claude Code and coding agents are used daily as a real part of the workflow. - Ship constantly. They deploy to production multiple times per day. Scope stays small and tight. - Remote, async. Communication happens via Slack and GitHub. No standups or status meetings for the sake of meetings. 🖼️ ABOUT THE ROLE Our hiring partner is looking for its founding engineer on the Creative Engine, and they're hiring data-first, not app-first. This is a data product at its core, and you'll own the foundation: the pipelines that pull and structure performance data across hundreds of accounts, the LLM workflows that analyze every video they run, and the search layer that turns all of it into answers strategists can query. You'll own features end to end, including the UI, but you don't need to be a frontend specialist. The heart of this role is the data and intelligence layer — you'll lean on Claude Code and agentic tooling to ship the dashboards and interfaces on top. Get the foundation right and the product follows, so they want someone who's genuinely LLM-fluent and savvy with AI-native development, not just someone who's heard of it. 🔨 WHAT YOU'LL DO - Design the data foundation: pipelines that pull, store, and structure performance, sales, and engagement data across hundreds of client accounts (performance creative, TikTok Shop, TikTok LIVE, and Meta), handling the messiness of real platform APIs. - Build the LLM analysis layer, using Gemini to watch and tag every video and Claude for generation, with structured output at scale and sane cost controls. - Build search and discovery with semantic search via pgvector, so a strategist can ask "what are the top hooks for beauty this week?" and get a real answer, not a keyword dump. - Build the "brain": connect viral organic references, top-performing ads, content tags, and creator history into queryable creative intelligence, accessible through the web app, Slack, and Cowork. - Develop creator-performance matching, linking creator history, content tags, and ad metrics to sharpen casting decisions. - Own the full stack on top of your data work: dashboards, tag browsers, enrichment queues, and export tools in Django + HTMX + Tailwind, moving fast with Claude Code (no dedicated frontend hire required). - Integrate external APIs (TikTok, Meta, Motion, Cruva) and internal async pipelines (Celery + Redis). THE STACK - Backend: Django, PostgreSQL, Celery + Redis, Django REST Framework - Frontend: Django templates, HTMX, Tailwind CSS (server-rendered, no SPA) - AI/ML: Google Gemini (video analysis), Anthropic Claude (script generation), pgvector (planned) - Infrastructure: Docker, AWS ECS/Fargate, Backblaze B2, GitHub Actions CI/CD - Integrations: TikTok API, Meta API, MagicBrief, Cruva, PayPal, DocuSign, Twilio, Slack 🔎 WHAT WE'RE LOOKING FOR Required: - 5+ years building production, data-heavy applications with Python/Django and PostgreSQL (senior / lead level, autonomous, low-process). - Strong SQL well beyond basic ORM queries (aggregations, joins, window functions) and comfort designing schemas and data models straight from use cases. - Experience designing and debugging async data pipelines (Celery + Redis or equivalent). - Genuinely LLM-fluent: hands-on with LLM APIs in production, prompt engineering, structured output, and cost management. - Fluent with agentic coding tools (Claude Code, Cowork) and excited to use them to move fast, including building full-stack on top of your own data work. - Experience integrating messy third-party APIs (rate limits, auth flows, inconsistent data). - Comfortable owning frontend (Django templates + HTMX + Tailwind) with AI assistance — you ship the whole feature, you just don't need to be a frontend expert. - Strong written communication and high ownership on a small, async team. Preferred: