Platform Engineer

Role Overview

Core Responsibilities

• Own and evolve the AI Factory’s core platform infrastructure: model serving, vector database infrastructure, embedding pipelines, and the compute layer that AI workloads run on.
• Design, build, and operate GPU-based inference infrastructure — including serving systems such as vLLM, TGI, or TensorRT-LLM — ensuring low-latency, high-throughput, and cost-efficient model serving at production scale.
• Build and maintain the data infrastructure that supports AI workloads: document ingestion pipelines, vector stores (pgvector, Pinecone, Qdrant), object storage, and the ETL processes that keep them current and reliable.
• Design and operate the deployment platform — containerised services, CI/CD pipelines, infrastructure-as-code — so that Engineers can ship to production quickly and safely.
• Own observability across the AI stack: define and implement telemetry, logging, tracing, and alerting for AI services, with specific attention to LLM-specific metrics such as latency distributions, token throughput, error rates, and model behaviour drift.
• Build shared internal tooling and platform abstractions that raise the productivity of engineers — reusable deployment patterns, model-serving clients, pipeline SDKs, and developer environment tooling.
• Set and enforce platform standards for reliability, security, and scalability across the AI Factory’s engineering teams.
• Partner with Engineers to understand their infrastructure requirements and translate them into durable platform capabilities, not one-off solutions.
• Lead technical decisions on infrastructure architecture, evaluate new platform technologies, and ensure the team is building on the right foundations for the next two to three years.

Basic Qualifications

• 10+ years of experience in platform, infrastructure, or backend engineering, with a demonstrable track record of building and operating production systems at scale.
• Deep experience with cloud infrastructure — particularly Azure, with AWS or GCP also valued — including compute, networking, storage, managed services, and cost optimisation.
• Strong expertise in containerisation and orchestration: Docker, Kubernetes, and the operational patterns that make containerised AI workloads reliable in production.
• Proven experience building and operating CI/CD pipelines and infrastructure-as-code — treating infrastructure as software, with the same standards for testing, review, and observability.
• Experience designing and operating data pipelines at production scale: ingestion, transformation, storage, and the reliability patterns that keep them running under real load.
• Strong proficiency in at least one backend systems language — Python, Java, or Go — with the ability to read and reason about code across the stack.
• Experience with relational databases at production scale, particularly PostgreSQL — schema design, query optimisation, connection pooling, and operational management.
• Hands-on experience with observability tooling — distributed tracing, structured logging, metrics, and alerting — and the ability to instrument systems so that failures are diagnosed quickly.
• Strong written and verbal communication — you can explain infrastructure decisions, failure modes, and trade-offs clearly to both engineering peers and non-technical stakeholders.

Preferred Qualifications

• Direct experience operating GPU infrastructure and LLM inference serving systems such as vLLM, TGI (Text Generation Inference), TensorRT-LLM, or equivalent — including performance tuning, batching strategies, and cost management.
• Experience building or operating vector database infrastructure — pgvector, Pinecone, Qdrant, Weaviate, or similar — at production scale, including indexing strategies, query performance, and operational maintenance.
• Familiarity with the infrastructure demands specific to RAG systems: document ingestion pipelines, embedding generation at scale, retrieval latency, and the operational complexity of keeping vector indices current.
• Experience building internal developer platforms or platform-as-a-product — with an understanding of what makes platform tooling actually adopted versus ignored.
• Familiarity with AI/ML workload patterns: batch inference, streaming inference, model versioning, and the infrastructure trade-offs involved in serving large models reliably.
• Experience operating real-time or streaming data infrastructure relevant to conversational AI or voice pipelines.
• Security and compliance experience in a regulated or government-adjacent environment.

How We Work

• **Ownership.**
• Takes full responsibility for the platform from design through operation — not to the boundary of a ticket.
• When engineers are blocked by infrastructure, this role unblocks them.
• **Force multiplier.**
• Measures success by the productivity and reliability it enables in others, not just the systems it directly builds.
• The best platform work is invisible to those who depend on it.
• **Evidence-driven.**
• Grounds every infrastructure decision in evidence: load profiles, failure data, cost analysis, and real operational experience.
• Avoids over-engineering and premature abstraction.
• **Bias for operability.**
• Builds systems that are easy to understand, debug, and operate — not just systems that work when conditions are ideal.
• Runbooks, alerts, and dashboards are first-class deliverables.
• **Closes the loop.**
• Tracks whether platform changes improved the outcomes that matter — latency, reliability, developer velocity — and iterates until they do.
• **Clear thinker.**
• Writes and communicates clearly.
• Can explain infrastructure decisions, failure modes, and operational trade-offs to engineers who do not share their infrastructure background.

Technical Depth Expectations

• Candidates will be expected to demonstrate genuine depth in at least three of the following areas.
• Conceptual familiarity is not sufficient.
• LLM inference infrastructure — GPU serving systems, batching strategies, latency and throughput optimisation, cost management, and model versioning in production.
• Vector infrastructure and retrieval pipelines — vector database operations, indexing strategies, embedding pipeline architecture, and retrieval performance at scale.
• Cloud infrastructure at scale — compute, networking, storage, managed services, and cost optimisation on Azure, AWS, or GCP, with a strong understanding of the trade-offs between managed and self-operated infrastructure.
• Platform and developer tooling — CI/CD, infrastructure-as-code, containerisation, and the design principles that make internal platforms actually useful to the engineers who depend on them.
• Observability for AI systems — distributed tracing, structured logging, metrics design, alerting, and the specific telemetry challenges of non-deterministic AI workloads.
• Data pipeline engineering — ingestion, transformation, reliability patterns, and the operational complexity of keeping AI data infrastructure current and trustworthy.