AI engineer in speech intelligence

HappyScribe; Barcelona; Híbrido; EUR 90.000 - 140.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Hybrid set-up

Flexible working hours

Barcelona office (Gràcia)

International & English-fluent

Competitive salary

Equity

Generous token spend policy

30 days holiday allowance

Health coverage

Tax-smart savings

Descripción de la vacante

HappyScribe in Barcelona is seeking an experienced AI-focused software engineer to own features end-to-end—from problem discovery to production, design prompts, and optimize retrieval and memory modules.

You will build semantic search across transcripts using embeddings and hybrid retrieval, ship with designers and engineers, and drive evaluation pipelines to catch quality regressions as models evolve.

Formación

• 5+ years as a software engineer with 2+ years in AI applications delivering real features
• Shipped LLM-based features to production with real users, not demos
• Fluent in prompt engineering, RAG architectures, and evaluation frameworks
• Experience shipping AI features in B2B SaaS; multi-provider LLM systems is a plus
• Experience with speech or audio AI is a plus

Responsabilidades

• Own AI features end-to-end—from discovery to production
• Build systems that turn conversations into intelligence: summaries, search, actions, memory
• Design evaluation frameworks to catch regressions and iterate as models evolve
• Optimize AI infrastructure for speed, cost, and reliability
• Ship features with product, design and engineering teams; build semantic search using embeddings

Conocimientos

LLM features in prod

Prompt engineering

RAG architectures

Evaluation frameworks

Herramientas

Embeddings

Hybrid retrieval

Semantic search

Descripción del empleo

Описание

Hello We’re HappyScribe, a profitable, 100% bootstrapped AI company based in Barcelona. We built one of the web’s most accurate multilingual transcription platforms, trusted by hundreds of thousands of users each month. That product funds our freedom to create what’s next—no investors, no burn rate, just focus. Now we’re taking that same tech and expertise to a bigger frontier: turning conversations into intelligence. We’re building the layer where speech becomes memory, insights become action, and teams work with more clarity and speed.

Задачи

• Own AI features end-to-end — from problem discovery to prompt design to production.
• You’ll build the systems that turn conversations into intelligence: summaries, search, actions, memory.
• Build and maintain our evaluation framework.
• You’ll design evals that actually catch regressions, iterate on prompts with confidence, and keep quality high as models change under us.
• Optimize our AI infrastructure for speed, cost, and reliability.
• Provider routing, caching strategies, chunking approaches, latency budgets — you’ll own the trade-offs.
• Stay dangerously current.
• The field moves weekly.
• You’ll bring new techniques to the team and know when a new model or approach is worth the migration cost.
• Ship with the product team.
• You’ll work closely with designers and engineers to integrate AI features that feel native, not bolted on.
• Design and ship semantic search across transcripts using embeddings and hybrid retrieval.
• Build an agentic system that turns meeting notes into drafted emails, tasks, and follow-ups.
• Create a prompt evaluation pipeline that catches quality regressions before users do.

Требования

• You’re a seasoned engineer (5+ years) who’s spent the last 2+ years deep in AI applications. You’ve shipped LLM-based features to production — not demos, not notebooks, real products with real users. You stay current by doing, not just reading. New model drops Tuesday, you've tested it by Thursday, and you know whether it's worth a migration or just hype. You're fluent in prompt engineering, RAG architectures, and evaluation frameworks. You've debugged retrieval that returned garbage and prompts that hallucinated confidently. You think in trade-offs: latency vs. quality, cost vs. capability, speed-to-ship vs. technical debt. You make calls and move. You're product-minded. You care more about solving user problems than implementing cool tech. You'll kill a feature that doesn't work. You are excited about our engineering principles and most of the Rails doctrine clicks with your style of work.
• Будет плюсом: You’ve shipped AI features in B2B SaaS where reliability and accuracy actually matter.
• Будет плюсом: You’ve built multi-provider LLM systems with fallbacks, routing, and cost optimization.
• Будет плюсом: You’ve designed retrieval systems at scale — embeddings, reranking, hybrid search.
• Будет плюсом: You have experience with speech or audio AI — transcription, diarization, or real-time processing.

Условия

A company culture built on trust, autonomy, and independence

• Hybrid set-up - Our setup is hybrid and built on trust.
• Flexible working hours — Work on the schedule that makes you the happiest and productive.
• Barcelona office (Gràcia) — More than a desk. A creative hub. Your space for collaboration and community.
• International & English-fluent — Global team, fully inclusive conversations.
• Competitive salary — our salaries are in the top percentiles of the Barcelona startup scene.
• Equity — By joining the company when the team is so small, you’re helping us set the foundations of our future success.
• Generous token spend policy — We expect our team members to work at the frontier of what's possible with AI. We expect spending, not saving, at the moment.
• 30 days holiday allowance — Recharge whenever you need to.
• Health coverage — Alan partnership covers half your medical insurance, anywhere you are.
• Tax-smart savings — Cobee by Pluxee for meals, transport, nursery. Keep more of what you earn.

We're asking you to come on an adventure, so if we find a treasure, we want you to get a piece.