SnowHeap LLC

SnowHeap is a UAE-based AI consultancy. We ship production LLM systems for regulated industries across MENA/EU/US and need an AI Compliance Engineer to own model governance, safety, and evaluations across our stack (OpenAI, Claude, Agno, CrewAI, LangChain/LangGraph/LangSmith). You’ll design controls that let us move fast and stay auditable: policy to pipeline, document to dashboard. Location: Remote (MENA/EU time zones) with optional Dubai meetups. Contract Type: 6-month temporary contract (with potential extension based on project needs and performance). Location: Remote (MENA/EU time zones). Candidates must be able to align at least 80% of their working hours with UAE time (depending on their location). Optional Dubai meetups. Please note: We are unable to provide visa sponsorship for this role at this time. Tasks Define and run SnowHeap’s AI governance program: policies, control library, risk register, exception handling, and sign-offs (from ideation to production). Map laws and frameworks (EU AI Act, GDPR/PDPL/DIFC DPL, NIST AI RMF, ISO/IEC 42001 & 27001, SOC 2) to concrete technical controls in our products and client projects. Build an evaluation harness for LLMs/agents: golden sets, scenario tests, adversarial probes, offline evals, and online A/Bs; track hallucination, safety, bias, privacy leakage, robustness, cost, and latency. Implement guardrails (PII detection, jailbreak/prompt-injection defenses, output filters, content safety) and wire them into pipelines (LangChain/LangGraph, CrewAI/Agno). Stand up audit-ready telemetry: data lineage, prompt/response logging with redaction, model cards, decision traces, and approval workflows (LangSmith/observability tools). Partner with Security/Privacy on DPIAs/TRA, retention, DLP, key management, access controls, and vendor risk (OpenAI/Anthropic terms, Azure/GCP/AWS). Lead red-teaming exercises; coordinate incident response playbooks for model failures and safety regressions. Review prompts, fine-tunes, and datasets for policy compliance; curate evaluation datasets and “go/no-go” acceptance criteria. Coach engineers, sales, and clients; write crisp docs and checklists; run internal trainings and readiness reviews. Contribute to proposals and client audits; turn compliance into a product advantage. Requirements 4+ years in Security/Privacy/Compliance, ML governance, or safety engineering, with 2+ years on LLM products. Strong grasp of LLM stacks: OpenAI & Azure OpenAI, Claude, Agno, CrewAI, LangChain/LangGraph/LangSmith. Hands-on model evaluation: building test sets, rubric-based scoring, offline/online evals, statistical analysis; familiarity with tools or libraries for evals/observability. Working knowledge of privacy & AI risk (GDPR/PDPL/DIFC DPL, EU AI Act concepts, NIST AI RMF), and how to turn them into safeguards, SOPs, and controls. Context engineering expertise: ability to design, test, and audit prompt chains, context windows, and memory architectures for compliance, safety, and explainability. Solid scripting in Python/Pydantic (TypeScript nice to have); able to review PRs and add compliance checks to CI/CD. Cloud/MLOps fluency: one of AWS/GCP/Azure; containers, secrets, monitoring, access controls. Excellent writing and stakeholder skills; can say “no” with rationale and ship a safer “yes”. Nice-to-haves ISO 27001/ISO 42001/SOC 2 implementation or audit experience. Prior red-teaming of LLMs (prompt-injection, data exfiltration, harmful content). Experience in regulated domains (financial services, healthcare, public sector). Arabic or UAE market experience. Benefits High-ownership role shaping SnowHeap’s AI governance and product roadmap. Remote-first across MENA/EU time zones; flexible hours. Competitive compensation with performance bonus. Fast career growth, build the function and lead it. If you’ve shipped real LLM systems and can turn regulation into running code and measurable quality, we want to talk. Apply with your CV plus any of the following (optional but helpful): Links to eval frameworks, safety work, or red-team write-ups you’ve done Example policies/checklists you authored (redacted is fine) GitHub or snippets showing eval harnesses, guardrails, or LangSmith/LangGraph workflows SnowHeap is an equal-opportunity employer. We value diverse perspectives and hire for talent, integrity, and impact.