Forward Deployed Engineer, requirement 9 of 12
Prototype to production
I take a prototype through generated test suites and merge review gates to a zero-downtime deploy that rolls back on failure.
The requirementDrive a solution from prototype or MVP, through testing, to a production deployment people depend on.
How I deliver it at a customer
- Build on the gatewayI build the prototype on the Luca Express multi-provider model gateway. Provider keys, rate limits, and per-call logging are in place from the first run.
- Generate the test matrixI run the platform test generator, which executes the T1 through T8 suites. It also generates RBAC, data isolation, and input fuzzing cases for every registered app.
- Gate the mergeI keep merges behind a scheduled gate that needs two distinct non-author approvals and no unresolved change request. Anything GitHub reports as not mergeable stays held.
- Check the release recordBefore any image swap I run the release approval check by hand. It exits non-zero unless an approved, not yet deployed release record exists for that commit.
- Smoke test the new podI also run six checks inside the new pod, from health and database readiness to one live model round trip. A failure holds the release and pages me rather than rolling back, because the gate runs outside the rollout.
- Roll out and watchI roll the new gateway image with maximum unavailable pods set to zero, so no request drops during the swap. The runner undoes any rollout that fails or times out. I then watch gateway health and the inference SLOs on Grafana dashboards and Loki queries.
Tools I use
| Tool | What it does here | Status |
|---|---|---|
| Platform test generator | Runs the T1 through T8 platform suites and generates RBAC, data isolation, and input fuzzing cases for every registered app. | Available |
| Scheduled merge gate | Merges a pull request only with two distinct non-author approvals, no unresolved change request, and a mergeable verdict from GitHub. | Available |
| Zero-downtime deploy runner | Rolls the new gateway image with zero unavailable pods and rolls back automatically when the rollout fails or times out. | Available |
| Release approval and smoke gates | One exits non-zero unless an approved release record exists for the commit; the other runs six checks inside the new pod. Neither is wired into the zero-downtime path yet. | Early access |
| Readiness rubric demo | Scores a prototype you describe against the same production gates listed on this page. | 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.
Describe a prototype in a sentence or two and this rubric scores it against the gates on this page. It returns a rating per gate and the specific gaps to close, not a pass or fail verdict.
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.AI workflow orchestrationI design the steps, contracts, and data flows, then ship them as running workflows on customer systems.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.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.