Forward Deployed Engineer, requirement 6 of 12
AI workflow orchestration
I design the steps, contracts, and data flows, then ship them as running workflows on customer systems.
The requirementDesign the AI workflows, APIs, integrations, orchestration, and data flows that connect a customer's systems to models.
How I deliver it at a customer
- Trace one real requestI sit with the operators and follow one live request through their systems, recording every hop and its owner. That written trace becomes the step list, so we agree scope before I write any code.
- Fix the API contractI publish the versioned HTTP contract first, the way the public API at api.onluca.ai/v1 serves integrations behind bearer keys. Each endpoint names the scope it requires, and the OpenAPI document at /v1/schema is generated from that contract.
- Encode the planI encode the flow as ordered steps in the job engine: API call, condition, loop, wait, log, notify, set variable, and abort. A set-variable step names the variable it writes, and an API call step stores its result in the shared response variable that later steps read.
- Bind tools and dataI expose each action as an agent tool on the tool bridge, and every tool declares the permissions it needs and its own timeout. A caller without those permissions is refused before the tool runs, and retrieval checks the session and the caller's document access before any search.
- Gate the writesI keep write steps behind an approval hold, so a job that changes infrastructure or migrates data waits for a person. I set retries, doubling backoff, and a timeout per job, and these holds are Early access, so I name the human owner.
- Ship and watchI deploy through ArgoCD onto Kubernetes and watch job duration and failure counts in Grafana and Loki. A failed smoke test or a breached service level objective rolls the release back.
Tools I use
| Tool | What it does here | Status |
|---|---|---|
| Agent runtime and tool bridge | Runs the agent loop and refuses any tool call that lacks the required permission, under a per-tool timeout. | Available |
| Public API, api.onluca.ai/v1 | Gives integrations a versioned HTTP surface with bearer keys, per-endpoint scopes, and a limit of 120 requests per minute. | Available |
| Multi-provider model gateway | Routes every model call through one gateway that holds the provider keys, applies per-user caps, and logs usage per call. | Available |
| Job engine and approval holds | Runs ordered multi-step plans with per-job retries, timeouts, and an approval hold; registered cron entries are not yet ticked by a scheduler. | Early access |
| Cloudflare Workers AI (Llama) | Plans and walks the steps behind the demo on this page, on public demo infrastructure. | Demo only |
Available means shipped and in use. Early access means built but not generally available. Roadmap means designed, not built. Demo only means it powers this public demo, not customer deployments.
The demo turns your request into an ordered plan, then walks the steps one at a time and times each one. It runs on public demo infrastructure and reads or writes nothing outside this page.
Open the live demoThe other answers
LLM applications in productionEvery model call goes through one gateway, guardrails run on both sides, and each task gets a measured model choice.Prompt engineeringI write the output shape first, version the prompt, budget its tokens, and compare model replies before shipping.Retrieval-augmented generationGrounded, cited answers over the customer's own documents, with access rules applied before retrieval.AI agents and orchestrationI build agents that call your systems as registered tools, and the workflows that trigger and sequence them.Request classification and routingI build automation that classifies each inbound request, routes it by role, and holds risky changes for human approval.API and integration designI publish the API contract first, gate it with keys and scopes, then build integrations against it.Hybrid enterprise integrationI connect AI services to your identity, applications and data, then run them where your policy allows.Prototype to productionI take a prototype through generated test suites and merge review gates to a zero-downtime deploy that rolls back on failure.Approvals and exception handlingI put a person between the agent and the change, with gated tools, a pending queue, expiry, and a record.Evals and output validationI score answers against a versioned rubric, trace failing calls to the integration that broke, and tune latency per call.Client-facing deliveryI turn engineering work into updates that client architects, security leads and business owners can each act on.
Gus IT LLC. Contact gus@gusit.de. Claude is a product of Anthropic PBC. Cloudflare and Workers AI are trademarks of Cloudflare, Inc. Llama is a trademark of Meta Platforms, Inc. Other names are trademarks of their respective owners. Gus IT LLC is an independent engineering firm and is not affiliated with or endorsed by these companies.