AI/ML Full Stack Engineer

Berkley Group; Cork; Híbrido; EUR 120.000 - 180.000; Jornada completa

Descripción de la vacante

Berkley Group in Ireland is seeking a Senior AI/ML Full Stack Engineer to lead production-ready AI systems. You will transform prototype AI concepts, including LLMs and agent workflows, into scalable microservices deployed with SLAs and robust observability.

You will own end-to-end development across the stack: FastAPI backends, React/Next.js frontends, and data stores (PostgreSQL + vector databases). The role emphasizes security, performance, and cost efficiency in hybrid work environments

Formación

• 5+ years in full-stack web software engineering, backend service architecture, and cloud operations.
• 2+ years dedicated to shipping LLM-driven applications, RAG pipelines, and autonomous agent workflows to live production users.
• Proven history establishing testing strategies (unit, integration, contract, load testing), prompt versioning, canary deployment flows, and incident response playbooks.

Responsabilidades

• Transform research AI prototypes into enterprise microservices with SLAs, automated fallback paths, and on-call resilience.
• Architect multi-step autonomous workflows using function calling, tool execution graphs, retries, timeouts, and circuit-breaker patterns.
• Develop streaming frontends and scalable APIs with React/Next.js and FastAPI, enabling end-to-end LLM-powered user workflows.

Conocimientos

Senior Software Engineer

Production AI & LLM

API Design

React/Next.js

FastAPI

PostgreSQL

CI/CD

Docker/Kubernetes

Cloud Infrastructure

OpenTelemetry

Herramientas

LangGraph

CrewAI

AutoGen

pgvector

Milvus

Weaviate

Descripción del empleo

Title: Senior AI/ML Full Stack Engineer

Target Location: Ireland (Hybrid - Cork, Limerick or Dublin)

Role Type: Permanent - Excellent Salary + Benefits - Hybrid

Expertise Level: Senior (5+ Years Software Engineering | 2+ Years Production AI & LLM Systems)

Candidate Profile Summary

A Senior Software Engineer specializing in transitioning prototype AI concepts-including LLMs, multi-agent frameworks, RAG architectures, and Text-to-SQL interfaces-into secure, scalable, and cost-efficient production applications. Experience covers the complete software lifecycle: high-throughput data ingestion, vector retrieval optimization, agentic orchestration, API design, responsive frontend interfaces, CI/CD automation, and full-stack LLMOps observability.

Primary Expertise & Deliverables

Production AI Engineering & Agentic Workflows

• POC-to-Production Conversion: Experienced in taking research-grade AI prototypes and engineering them into enterprise-ready microservices with strict SLAs, automated fallback paths, and on-call resilience.
• Agent Orchestration: Architecting multi-step autonomous workflows using function calling, tool execution graphs, dynamic retries, timeouts, and circuit-breaker patterns.
• Advanced RAG Pipelines: Designing end-to-end retrieval frameworks-including adaptive document chunking, metadata filtering, hybrid search index strategies, cross-encoder re-ranking, and grounded generation.
• Performance & Cost Optimization: Implementing semantic caching, response caching, token usage budgets, and model routing to lower query latency and operational overhead.

Full-Stack & API Architecture

• Asynchronous Backends: Building typed FastAPI applications with OpenAPI specifications, Pydantic DTOs, async I/O, rate limiting, connection pooling, and idempotent background processing.
• Modern Streaming Frontends: Developing React and Next.js interfaces optimized for streaming LLM responses, multi-step agent status tracking, complex state management, and continuous user workflows.
• Relational & Vector Data: Extensive work with PostgreSQL, pgvector, standalone vector databases, and SQL query tuning.

LLMOps, Reliability & Security

• Evaluation & Testing: Building continuous evaluation gates (CI eval gates, custom golden datasets, automated regression suites, trace replays) to prevent quality degradation prior to release.
• Full-Stack Observability: Instrumenting applications with OpenTelemetry to track token burn rates, cache hit ratios, end-to-end latency traces, and custom SLO/SLI metrics.
• AI Security & Governance: Applying defense-in-depth strategies against prompt injection, data leakage, and unauthorized tool invocation, while maintaining RBAC/ABAC and audit readiness.
• DevOps & Infrastructure: Containerizing workloads via Docker and Kubernetes, managing cloud infrastructure (AWS/Azure/GCP) using Terraform, and building automated GitHub Actions pipelines.

Core Technical Matrix

Domain

Key Technologies & Frameworks

AI / LLM Frameworks

RAG, Agentic Tool Chains, Multi-Agent Orchestration (LangGraph / CrewAI / AutoGen), Text-to-SQL, Structured Generation (Pydantic / JSON Schema)

Retrieval & Vector Search

Hybrid Search (Dense + BM25), Cross-Encoders, Semantic Caching, Re-ranking, pgvector, Milvus, Weaviate

Backend & Databases

Python 3.10+, FastAPI (Async), PostgreSQL, Task Queues (Celery / RQ / Arq), SSE, WebSockets, REST / OpenAPI

Frontend & UI

React, Next.js, TypeScript, React Query, Zustand, Optimistic UI Updates, Streaming UI Components

DevOps & Cloud

Kubernetes, Docker, Terraform (IaC), GitHub Actions, Cloud Storage (S3 / GCS / Azure Blob), AWS / GCP / Azure

LLMOps & Observability

OpenTelemetry, Prometheus, Grafana, CloudWatch, Evaluation Frameworks (RAGAS / TruLens), Golden Set Testing

Security & Compliance

Prompt Injection Defense, DLP, Secrets Management (Vault / Cloud KMS), RBAC/ABAC, SOC2/ISO Alignment

Experience & Track Record

• 5+ years in full-stack web software engineering, backend service architecture, and cloud operations.
• 2+ years dedicated to shipping LLM-driven applications, RAG pipelines, and autonomous agent workflows to live production users.
• Proven history establishing testing strategies (unit, integration, contract, load testing), prompt versioning, canary deployment flows, and incident response playbooks.