Initializing quality gate
click anywhere or press any key
Forward Deployed Engineer — AI Agents & LLM Orchestration

Filipe Pawlik Leite

I make AI agents stop lying.

I build the harness around AI agents — the gates, evaluation and orchestration that decide whether an agent's output is good enough to ship. Then I sit with the customer, find the real problem, and stay until the thing works.

Watch the gate refuse
02 — The thing I actually sell

A gate that
refuses.

Anything an agent writes goes through a gate before shipping: typecheck, a real test suite and — on web — a build that produces a servable artifact. Nothing ships otherwise. Flip the switch and break the code. The gate below is the same logic, running here in your browser.

agent-output.ts 

      

      

        
Quality gateidle

        
    
        
    
          
    
            Ready
            
    Press run. The gate does not know which variant you picked.
    
          
    
          
        
    
      
    
    
    
    
    Proven honest with shadow tests: known-bad code is refused in front of whoever opens this page. When the gate fails, the error goes back to the agent, which retries with backoff — up to a hard cap.
    
  
    

    


    


    
  
    
    
    03 — Measured, not asserted
    
    
    Every number here
    was counted.
    
    
    If a figure on this page has no source, it should not be on this page. Small samples travel with their sample size.
    
    
    
  
    

    


    
  
    
    
    04 — Eight agents, one router
    
    
    Routing by hit rate,
    not by vibe.
    
    
    Measured accuracy decides who gets the next task, sample size attached. Pick a job and watch which agent the router would call, and why — the same rule it runs in production.
    
    
    
      
    
      
    
        
    
        
    
        
    A hit rate is only shown where n is large enough. The rest appear with their n — too few runs to route on.
    
      
    
    
    
  
    

    


    
  
    
    
    
      
    05 — Twenty years
    
      
    Hospitals, blood banks,
    governments, Chevrolet.
    
    
    
    
    
    
    
  
    

    


    
  
    
    
    06 — What I build
    
    
    Systems that run
    without me.
    
    
    
  
    

    


    
  
    
    
    07 — How I pick technology
    
    
    For every layer, the same three questions: what it really costs, how good it actually is, and whether it holds when the volume grows. Then I measure — and the decision is what survived measurement, not the favorite candidate.
    
    
    
  
    

    


    
  
    
    Open to Forward Deployed roles
    
    Let's find your
    real problem.
    
    filipe@netcks.com
    
    Bombinhas, Santa Catarina, Brazil. UTC−3 — full overlap with US business hours. English, Spanish and Portuguese, all fluent, all working languages.