Software Engineer, AI / Backend Systems: build dependable AI systems.
CiteSurge is hiring one full-time, globally remote engineer. This is applied research and experimental development across non-deterministic AI evidence, backend services, data systems, and multi-provider infrastructure.
The work
CiteSurge measures how brands appear in AI-generated answers and turns that evidence into reliable product behaviour. The hard part is not calling a model. It is extracting dependable signal from outputs that are non-deterministic, differently structured across providers, and liable to change without notice. A competent engineer cannot derive every correct approach from established documentation; the unresolved parts have to be tested systematically against known cases and retained evidence.
You will design, test, and validate inference methods, data models, provider adapters, queues, backend services, and deployment paths. The role joins applied AI work with the platform engineering required to make experiments reproducible and safe enough for customer-facing use. You should be comfortable moving between a failed evidence case, a typed domain model, a worker or queue, an API boundary, and an operational dashboard.
Research and engineering problems
Multi-engine citation evidence
ChatGPT, Claude, Gemini, Perplexity, Google AI experiences, Bing Copilot, Grok, and future providers expose answers, citations, and mentions in different forms. You will improve the adapter and normalization layer, validate response-shape changes, preserve inspectable evidence, and help distinguish a genuine no-citation answer from an incomplete collection. Each provider is treated as an empirical integration: probe real behaviour, validate against fixtures and known cases, then wire the result into the pipeline with neutral failure states.
Entity intelligence
A token such as “Pro” or “Lab” may identify the customer, an unrelated organization, a product, or ordinary language. You will work on a tiered entity system that combines aliases, neural named-entity candidate detection, deterministic scoring, collision memory, and model-assisted arbitration for uncertain matches. The goal is not a clever demo. It is an evidence trail that improves precision without silently discarding difficult cases.
Citability and causal uncertainty
CiteSurge evaluates whether passages are clear, self-contained, supported, and likely to be useful as sources, then compares that analysis with historical citation evidence. You will help calibrate deterministic citability signals against an evidence corpus while keeping prediction separate from proof. Later answer-engine movement can be observed; it must not automatically be presented as caused by one edit.
Platform infrastructure
You will build provider adapters, queues, cost attribution against versioned rate cards, backend services, drift and readiness checks, compatibility validation, and deployment systems. The product has to remain useful when a provider is unavailable, a response changes shape, an integration is only partially configured, or a target environment behaves differently from a local fixture.
What we are looking for
- Strong backend engineering in TypeScript, Go, Python, or a comparable systems language.
- Experience designing APIs, queues, data models, workers, integrations, and observable production services.
- Practical work with LLMs, information extraction, retrieval, evaluation, entity resolution, or probabilistic systems.
- A habit of turning uncertainty into hypotheses, fixtures, experiments, and written findings.
- Good judgment about customer-safe failure states, privacy boundaries, security, and proprietary implementation detail.
- A body of work you can explain line by line, including decisions that did not work and what changed afterward.
What we offer
- Full-time, globally remote work with documented handoffs and agreed collaboration overlap.
- Competitive salary, transparent bands, and equity after one year.
- A four-week summer slowdown and remote-first working practices.
- Meaningful ownership in a small team working on unsolved applied-AI and infrastructure problems.
- Room to publish general technical learning while protecting customer data and proprietary methods.
How to apply
Email a short note, your CV or profile, and two examples of systems work you are proud of. Describe one ambiguous or non-deterministic technical problem you investigated and how you determined whether the result was reliable. A concise explanation is more useful than a long cover letter.