AI Architect

Description

The company operates in 149 countries and has more than 370,000 people working across audit and assurance, tax and legal, and deals and consulting.

As an AI Architect, you will take production Generative AI solutions for clients from first discovery through to delivery. The role combines hands-on architecture, technical leadership and client work. You will decide where Generative AI creates value, build secure and scalable systems, and track their quality and cost once they are live.

Responsibilities

• Architecture and delivery: lead the design and delivery of Generative AI solutions and guide development teams on suitable tools, frameworks and platforms.
• Turn business requirements into solid application, service and target-state architectures, including reusable patterns for RAG, agent orchestration, prompts and evaluation.
• Choose the right adaptation approach for each use case (prompting, RAG, fine-tuning or distillation), and recommend against Generative AI when it is not the right fit.
• Own scalability, performance, latency, observability, security and cost across AI solutions.
• Lead technical design and code reviews, and keep delivery aligned with client needs and quality standards.
• Quality, security and operations: own the evaluation strategy, including golden datasets, offline and online evaluation, human-calibrated LLM-as-judge and CI regression tests.
• Set retrieval quality standards for chunking, hybrid search, reranking, query rewriting and permission-aware retrieval, measured by recall and precision.
• Define controls for AI security, privacy, Responsible AI and governance, covering prompt injection, data exfiltration, jailbreaks, sandboxing, least-privilege identity and supply-chain risk.
• Manage solution economics through model routing, caching, batching, capacity planning and defensible cost-per-transaction estimates.
• Support enterprise architecture standards and help assess technology options.
• Leadership, clients and practice: supervise vendor developers, mentor engineers and junior architects, and improve internal design and development practices and documentation.
• Support business development by finding and researching opportunities with new and existing clients.
• Build internal relationships, help strengthen the firm's reputation, and grow your skills in line with company and team priorities.
• Take on other responsibilities that fit the role and are agreed with your manager.

Requirements

• Experience: at least 5 years of agile product delivery and hands-on experience designing and delivering production software at scale.
• Demonstrated expertise in Generative AI solution architecture, including having personally put at least one Generative AI system into production and being able to explain its architecture, failure modes, operating costs and lessons learned.
• Bachelor's or Master's degree in Computer Science or Engineering, or equivalent practical experience.
• An Azure certification such as Azure Solutions Architect Expert or Azure Developer Associate is preferred.
• Strong hands-on Generative AI architecture experience is essential. You should have worked across several relevant technologies; experience with every named tool is not required.
• Large language models such as GPT-4, Gemini and Llama.
• RAG, agent orchestration and Microsoft Copilots, using platforms or frameworks such as Azure OpenAI Service, Semantic Kernel, LangChain, AWS Bedrock or Google Vertex AI.
• Vector search and retrieval engineering: chunking, hybrid search, reranking, query rewriting, metadata filtering and permission-aware retrieval.
• Production AI agents: tool calls, state and memory, durable execution, approval controls, retries, idempotency and cost safeguards.
• Structured generation: schema-enforced outputs, validation and repair, deterministic fallbacks and graceful degradation.
• Cloud-native Azure architecture, including Azure Machine Learning, Azure AI services and suitable infrastructure and platform services.
• Service-oriented, event-driven and microservices architectures.
• Relational and NoSQL data stores such as Microsoft SQL Server and MongoDB, with sound judgement on technology selection.
• Containers and orchestration with Docker and/or Kubernetes.
• DevOps and delivery: CI/CD, automation, environment management and tools such as Azure DevOps, GitHub or Jira.
• Working knowledge of user-centred product practice, including user stories, personas and prototyping.
• AI evaluation with golden datasets, LLM-as-judge and prompt and model regression testing.
• LLMOps and observability: tracing, token and cost attribution, prompt and model versioning, trace replay and quality-drift monitoring, using tools such as LangSmith, Langfuse or Azure Monitor.
• Model economics: routing, cascading, caching, batching and provisioned versus consumption capacity planning.
• AI security: prompt injection, data exfiltration, jailbreak resistance, sandboxing, least-privilege identities, supply-chain risk and the OWASP Top 10 for LLM Applications.
• Model Context Protocol and agent interoperability standards.
• Model adaptation beyond prompting and RAG, including fine-tuning, LoRA/PEFT and distillation.
• Multimodal AI, including document intelligence, vision, speech or voice agents.
• Knowledge of AI regulation and assurance frameworks, including the EU AI Act, ISO/IEC 42001 and NIST AI RMF.
• Consulting or professional-services experience with external clients.
• Strategic, analytical problem-solving with close attention to detail, plus leadership, troubleshooting, curiosity and creativity.
• Fluent written and spoken English, with clear communication with product teams, clients and stakeholders.
• Strong collaboration and client focus: understanding needs, acting on feedback and supporting timely, high-quality delivery.
• Ongoing learning about emerging technology trends.
• Good judgement on when Generative AI adds value and when a simpler solution is better.

Conditions

Hybrid working, with a modern office in central Sofia, Bulgaria.

The hiring process has three steps: a case study, a technical interview and a final interview.

Personal data in your application is controlled by the employing firm and processed for this recruitment, and kept for future recruitment or job event invitations only with your explicit consent.

Benefits

• Work with leading local and international clients, industry experts and the global network.
• Challenging, meaningful projects that help you grow professionally and make an impact.
• Technical training, coaching and support for relevant professional qualifications.
• 25 days of annual leave, additional health insurance, meal vouchers at the maximum amount, a Multisport card, annual fringe benefits and wellbeing initiatives.
• Clear career growth paths and support for a long-term career.