Forward Deployed Engineer, requirement 4 of 12
AI agents and orchestration
I build agents that call your systems as registered tools, and the workflows that trigger and sequence them.
The requirementBuild agentic AI solutions, with hands-on experience in AI agents and workflow orchestration.
How I deliver it at a customer
- Scope the jobI write down the task, the systems it touches, and the actions the agent may take. That list becomes the tool catalog the tool bridge serves at GET /tools.
- Wire the toolsI register each action in the tool bridge with a JSON schema and a permission list. Calls run through POST /tools/:name/execute, which checks the caller's permissions and refuses a gated tool the caller does not hold.
- Stand up the agentI open the session through POST /agents/create in the agent runtime. The runtime records each session, tool call and token count in Postgres, so a session reloads after a restart.
- Set the guardrailsI turn on the policy engine, which matches each command against a risk list and blocks a fixed destructive set. A dangerous command is held as an approval request and does not run until a person approves it. Every decision, approved or rejected, is written to the audit trail. This part is built but not yet generally available.
- Orchestrate the workflowI wire the CRM sales automation engine to fire on events such as lead created and deal stage changed. It runs the workflow's ordered steps, evaluates the condition on each step, and persists every step result. Delayed follow-ups are queued, but nothing runs that queue on a schedule yet.
- Record what ranI check that every tool call, token count and workflow step result is written to Postgres. The agent and workflow endpoints read those records back, so the customer can see what ran. Pod logs reach the platform's Grafana and Loki stack; no agent-specific metric is exported yet.
Tools I use
| Tool | What it does here | Status |
|---|---|---|
| Agent runtime (/agents/*) | Creates agent sessions, runs the tool-calling loop, and records every tool call and token count. | Available |
| Tool bridge (/tools/*) | Registers each action as a tool with a schema and permissions, then executes calls against it. | Available |
| CRM sales automation engine | Fires event-triggered workflows whose ordered steps run with conditions and persisted results. | Available |
| Policy engine with Luca Gatekeeper | Classifies command risk, holds dangerous actions for human approval, applies role checks, and writes the audit trail. | Early access |
| Cloudflare Workers AI (Llama) demo agent | Runs the public tool-calling loop behind the demo 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.
Ask a question and the agent decides which of its three tools to call, and in what order, within a four-step limit. The page prints every call, its arguments and its result, then the final answer.
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.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.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.