Descripción de la oferta
Location: Barcelona, Spain , 0% remote. 100% in-office.
Consulte la descripción del puesto a continuación. Si confía en que tiene las habilidades y la experiencia adecuadas, envíe su solicitud hoy mismo.
Employment Type: Full-time
Level: Senior (Reporting to: Founding team / CPO and Head of AI)
AIMA ( ) is seeking an AI Engineer to design, build, and operate the core conversational and cognitive systems of our AI companion platform. The successful candidate will develop production-grade LLM-based capabilities including dialogue orchestration, memory and retrieval, safety guardrails (with particular attention to healthcare-adjacent use cases), and evaluation frameworks to ensure reliability over time.
This role requires a strong blend of applied AI engineering, systems thinking, and product sensibility. You will work closely with the founding team to translate product requirements into robust technical architectures and deliver measurable improvements in user experience, safety, and system performance.
Key Responsibilities
2.1 Conversational System Engineering
Design and implement a layered conversational architecture (policy, persona, memory, task execution).
Develop dialogue state management, conversation flow control, and escalation logic.
Improve response quality, coherence, and user alignment through iterative prompt and system design.
2.2 Memory, Retrieval, and Personalization
Build long-term memor components (episodic and semantic memory, user preferences).
Implement retrieval strategies including RAG and structured memory access patterns.
Define memory hygiene policies (relevance filtering, decay, correction, reinforcement) and maintain high integrity of stored information.
2.3 Agent Orchestration and Tool Use
Develop multi-agent or modular "organ-based" systems (e.g., planner, conversational engine, memory module, safety module).
Implement tool invocation, routing, arbitration, and failure-handling strategies.
Optimize latency, cost, and robustness across model calls and tool pipelines.
2.4 Safety, Guardrails, and Compliance-by-Design
Implement guardrails for sensitive topics, including healthcare-adjacent conversations.
Develop uncertainty handling, safe-completion patterns, refusal/deflection behaviors, and escalation to human guidance when appropriate.
Contribute to practices aligned with EU expectations on transparency, safety, and responsible AI.
2.5 Evaluation, Monitoring, and Quality Assurance
Create evaluation harnesses, golden test sets, and regression testing for conversational behavior.
Define measurable KPIs (e.g., helpfulness, hallucination rate, safety compliance, retention signals).
Establish monitoring, logging, and feedback loops to support continuous improvement.
2.6 Production Engineering
Deliver maintainable, tested, and well-documented code.
Support deployment and operation of AI services in production environments.
Collaborate with product and engineering stakeholders to prioritize and deliver roadmap items.
Required Qualifications
Strong software engineering experience in Python.
Hands‑on experience building and shipping applications using LLMs (hosted APIs and/or local/open‑source models).
Practical experience with RAG, embeddings, vector databases (e.g., Chroma, FAISS, or similar), and chunking/retrieval strategies.
Demonstrated ability to design and operate reliable systems (testing, monitoring, incident awareness).
Strong written communication and ability to explain technical decisions clearly.
Preferred Qualifications
Experience with multi‑agent architectures, state machines, or tool‑using LLM systems.
Familiarity with evaluation methodologies for generative systems (offline benchmarks, human‑in‑the‑loop review, regression suites).
Exposure to healthcare‑adjacent domains, safety‑critical product constraints, or regulated environments.
Experience with speech pipelines (ASR/TTS) or multimodal systems (future roadmap).
Background or interest in cognitive architectures, HCI, or long‑term human–AI interaction.
Core Competencies
Systems thinking and architectural judgment.
Product‑oriented engineering (balancing user value, safety, and technical constraints).
High ownership, autonomy, and comfort operating in ambiguity.
Strong quality bar for correctness, safety, and reliability.
Ability to collaborate effectively in a small, fast‑moving team.
100% on‑site work.
Salary Expectations: Starting at 45K€ per year, depending on value.
#J-18808-Ljbffr
