Senior ML/AI Engineer (LLM)

About the Role
We are hiring a Senior ML/AI Engineer to join a small, senior, fully dedicated team building an AI-powered content generation platform for a US-based client. The platform uses LLMs to generate rich, structured content — including interactive components and media — and integrates with the client's existing systems.

Content safety is a core constraint of the product, not an afterthought: every generation passes through guardrails that validate vocabulary, tone, and subject matter before anything reaches an end user. This role is where most of that work happens.

What You'll Do

1. Build the LLM generation pipeline: prompt construction from structured inputs, model calls, and structured output that maps to renderable content components

2. Design and implement safety filters and guardrails: content moderation, tone and vocabulary validation, and output checks before content is shown to users

3. Build output evaluation: eval sets, automated quality scoring, regression testing on prompt and model changes, and human-review workflows

4. Implement retrieval (RAG) so generated content is grounded in the client's reference material

5. Build and operate the AWS serverless services behind the pipeline: Lambda, Step Functions, API Gateway, SQS/EventBridge, DynamoDB/S3

6. Build the image generation pipeline for in-product graphics, deployed within the client's AWS environment (Bedrock image models or SageMaker-hosted alternatives); contribute to a longer-term generative video roadmap

7. Handle production LLM concerns: streaming, caching, retries, rate limits, cost monitoring, and observability

8. Collaborate daily with the team lead, full-stack engineers, and the client's internal teams

Required Qualifications

1. 5+ years of software engineering experience, with 1–2+ years shipping production features on top of LLMs (not just prototypes)

2. Strong Python; comfortable contributing in TypeScript/Node when the pipeline touches shared services

3. Hands-on experience with AWS Bedrock (model access, IAM, invoking hosted models in production) — experience with other LLM APIs (Anthropic, OpenAI, Google Vertex) is valuable but not a substitute

4. Strong grasp of function calling / structured outputs (JSON schema) for reliable, renderable generation

5. Production experience with image generation (Bedrock image models, Stable Diffusion, or similar), ideally deployed inside a client's cloud environment rather than via external APIs

6. Experience building AWS serverless architectures: Lambda, API Gateway, Step Functions, and event-driven patterns

7. Experience implementing LLM guardrails or content moderation: moderation APIs, safety classifiers, output validation, prompt-injection defenses

8. Experience with LLM evaluation: building eval sets, LLM-as-judge, regression testing prompt/model changes

9. Experience with RAG: embeddings, vector search, chunking, retrieval quality tuning

10. Advanced English — daily written and spoken communication with a US client team; attends client standups on US hours

Nice to Have

1. Fine-tuning experience (LoRA/PEFT, Bedrock custom models) or serving open-weight models (Hugging Face, vLLM, SageMaker endpoints)

2. Video generation experience (emerging video models or video-generation services) — relevant to the product's longer-term roadmap

3. LLM observability/orchestration tooling: LangSmith, Langfuse, LangChain/LlamaIndex

4. Enterprise security posture experience (IAM, least-privilege, sensitive data handling)

5. Experience with content-safety requirements for regulated or sensitive audiences

What We're Not Looking For

1. Classical ML specializations (computer vision, OCR, forecasting, recommenders) — this role is entirely LLM/generative AI focused

2. Research-only profiles — this is a product engineering role that ships to real users