Forward Deployed Engineering

About the Company

Cyber Sierra makes security compliance easy for enterprises using its AI enabled platform in driving continuous controls monitoring, third party risk management and governance. Our People are of high integrity, value good work ethics, and mission-focused.

About you

Mindset

You are a self-starter who can roll up your sleeves and hustle. You are driven to achieve the correct result through smart work and integrity. You are a motivated individual, ready to build from scratch and grow the impact of your work positively for users globally. You are not driven by hierarchy, your drive lies in your curiosity and pushing boundaries of technology.

Why Cyber Sierra

Cyber Sierra is a mission-driven organization, with the aim to help enterprises, globally, to conquer new age cyber risks and compliance needs. In the process, Cyber Sierra seeks to build the best cyber experiences for professionals. If you want to work with like-minded people and be part of a movement then come speak to us at [email protected].

What will you do?

Our highest-value deployments do not look like a SaaS signup. They look like an air-gapped environment inside a regulated bank, a government cloud tenancy where nothing can leave the network, or a security team that needs our AI's reasoning to hold up under a formal governance review. As a Forward Deployed Engineer, you are the person who makes those deployments work and build sharp insights that grow the technical relationship for better performance. You write the code, you own the evaluation harness, and you are accountable for whether the system works in the customer's environment.

1. Work directly with enterprise and public-sector customers to map their real compliance workflows, such as questionnaire response, evidence review, SSP verification and vendor risk assessment, into agentic systems that complete the work end to end.

2. Design and build agentic pipelines on our platform, covering tool-calling loops, multi-step document understanding, retrieval over messy evidence corpora, and the harness code that makes an agent reliable in production.

3. Own the evaluation story for everything you ship. Build golden datasets from customer ground truth, define metrics that track what customers actually care about, wire them into CI with hard pass thresholds, and hold the line when a change regresses them.

4. Deploy into constrained environments including on-prem, air-gapped and government cloud. This covers self-hosted open-weight models, quantised builds, GPU sizing and self-hostable observability. You will need to know what breaks when internet egress disappears, and know it before the customer does.

5. Write the documentation that gets us through technical scrutiny: governance responses, evaluation methodology write-ups, and architecture and dataflow documentation for customer security and AI risk reviewers.

6. Feed learning back into the roadmap. Every bespoke capability you build is a hypothesis about what the platform should do natively and you shall make the case.

What you need to know?

Agentic architectures. You have built and debugged more than a chat wrapper. You can reason about when a single-agent loop is the right shape and when the problem needs orchestration across several agents, and explain your reasoning. You understand context engineering, tool design, retrieval and document-understanding pipelines, memory and state management, and common failure modes such as loop divergence, context degradation over long runs and silent tool errors.

Eval-driven development. You treat evaluation as the development method and not a QA phase at the end. You have practical experience building offline eval harnesses, curating golden sets, working with reference-based and LLM-as-judge metrics such as RAGAS or equivalents, setting and defending CI thresholds, and instrumenting production traces to catch what offline evals miss. You understand the gap between a benchmark number and a claim that will survive a customer's technical reviewer.

Local and air-gapped deployments. Hands-on experience with self-hosted inference using vLLM, Ollama or similar, quantisation and its accuracy trade-offs, containerised deployment on Kubernetes, and self-hostable tracing and observability tooling. You are comfortable working somewhere with no outbound network and no managed services to fall back on.

Engineering fundamentals. Strong production Python or TypeScript, comfortable across the stack and into infrastructure. You debug from logs and traces.

Customer-facing judgement. You can sit in a room with a CISO, an auditor and a platform engineer, build on critical context, and be honest about product limitations as well as opportunities so that all of us can succeed together.

Nice to have

1. Experience with regulated or public-sector procurement and deployment constraints

2. AWS depth, particularly Bedrock and sovereign or government cloud environments

3. If you have enough curiosity that you play with air-gapped AI models or try to deploy in different environments, you’re a great fit already.