Desarrollador/a Fullstack

Tendios, Inc.; Barcelona; Hybrid; EUR 90,000 - 130,000; Full time

Job summary

Tendios, Inc. in Barcelona is seeking a Senior AI/ML Engineer to design the intelligence behind our SaaS platform for tender documentation.

You will decide models, prototype in Python, and ship features into our Node.js/NestJS/TypeScript stack, collaborating with the AI & Data squad. You will build AI architectures for extraction, classification and summarisation, manage cost and latency across commercial APIs and open-weight models, and ensure production-grade data pipelines and tests.

Qualifications

• 5+ years software engineering with 2+ years delivering production ML features

Responsibilities

• Design AI architectures for extraction, classification, matching, summarisation and workflows over tender documents
• Prototype in Python and ship features into a Node.js/NestJS/TypeScript platform
• Build retrieval pipelines with embeddings, hybrid search and reranking
• Manage cost, latency and quality per feature and per customer
• Ensure tests, observability and CI/CD for production ML features

Skills

Python for ML

PyTorch

TypeScript/Node.js

LLM APIs

Open-source models

RAG retrieval

Docker

CI/CD

English fluency

Education

Bachelor's in CS or related field

Tools

Langfuse

Job description

About Tendios

Tendios is a SaaS platform for Spanish public procurement intelligence. Our products, Bid and Create, help companies find, analyse and win public tenders. Tender documentation is long, unstructured and full of legal and technical detail, which makes it an ideal ground for applied AI.

The role

We are looking for a Senior AI/ML Engineer who designs the intelligence behind our products and ships it into production code. You will decide which model solves each problem, whether a commercial API, a fine-tuned open-source model or a combination, and make it fast, reliable and affordable at scale.

This is not a research-only position. You will prototype in Python, then integrate and ship features in our Node.js / NestJS / TypeScript platform alongside the rest of the engineering team. You will join our AI & Data squad.

What you'll do

AI architecture and models

• Design the AI systems behind our features: extraction, classification, matching, summarisation, drafting and agentic workflows over tender documentation.

• Choose the right approach per use case: prompting, RAG, fine-tuning, distillation or classic ML, and justify it on quality, latency and cost.

• Build retrieval pipelines: chunking, embeddings, hybrid search and reranking over large document collections.

Commercial LLMs, cost-efficient

• Own our usage of commercial providers (Anthropic, OpenAI, Google and others) and keep cost per feature under control.

• Apply prompt caching, batch APIs, model routing and cascading, semantic caching, context compression and token budgets.

• Track cost, latency and quality per feature and per customer, and set alerts before bills surprise us.

Open-source models

• Evaluate, fine-tune (LoRA / QLoRA) and deploy open-weight models such as Llama, Mistral, Qwen or Gemma where they beat APIs on cost, privacy or control.

• Serve them efficiently with tools like vLLM, TGI or llama.cpp, using quantisation and batching to get the most from each GPU.

• Build the data pipelines and labelled datasets needed for training and evaluation.

Product engineering

• Ship AI features end to end in our Node.js / NestJS / TypeScript platform: APIs, background jobs, streaming responses and integrations.

• Build evaluation suites and regression tests so model or prompt changes never silently degrade quality.

• Propose new AI-driven features with product, and help the team adopt good LLM engineering practices.

What you bring

• 5+ years of software engineering, including 2+ years building LLM or ML features that run in production.

• Strong Python for ML work: PyTorch, Hugging Face Transformers and the surrounding data tooling.

• Solid TypeScript and Node.js in production; NestJS experience or the ability to become productive in it quickly.

• Deep hands-on knowledge of commercial LLM APIs, including concrete examples of how you reduced their cost without losing quality.

• Practical experience fine-tuning, quantising and serving open-source models, and a clear view of when self-hosting pays off versus an API.

• RAG in production: embeddings, vector search (e.g. pgvector, Qdrant, OpenSearch), hybrid retrieval and reranking.

• An evaluation mindset: you build test sets, use LLM-as-judge carefully and measure before you claim improvements.

• Comfort with the full lifecycle: Docker, CI/CD, cloud infrastructure, observability and on-call ownership of what you ship.

• Fluent English for technical work.

Nice to have

• Spanish, since our source documents and many users work in Spanish.

• Agents and tool use: function calling, the Model Context Protocol (MCP) and multi-step workflows.

• Document AI: PDF parsing, OCR, layout and table extraction from messy files.

• LLM observability and gateway tools such as Langfuse, LiteLLM or similar.

• GPU infrastructure and cost management on AWS, GCP or specialised GPU providers.

• Knowledge of public procurement, legal or other document-heavy domains.

• Awareness of GDPR and the EU AI Act when designing AI features.

How we work

• Remote from anywhere in Spain, or hybrid from Barcelona.

• A 16-person engineering organisation in four squads; you join the AI & Data squad and work closely with product squads.

• Real ownership: you choose the tools and models, and you are measured on what reaches customers.

• Open question: salary range and benefits to add before publishing.