Unnati Shah — AI Engineer

~/unnati$ai engineer @ johnson & johnson

{{ termCard.tag }} {{ termCard.text }} click → {{ termCard.cta }}

I build the orchestration, retrieval, and evaluation layers behind production LLM systems.

// currently

Architecting multi-agent systems at Johnson & Johnson. Before that, Text-to-SQL at C3.ai and machine learning research at USC, published in the Strategic Management Journal.

↳ psst: click one of the little bots in the background. They run code.
running 

        

        

          

            
          

          

            
name
Unnati Shah

            
role
AI Engineer · Johnson & Johnson

            
based in
New York City

            
studied
M.S. CS · USC

          

        

        
        

      

    

  


  

    
  


  

    

      

        01 // Work experience
        
Systems in production

      

      
Case studies from Johnson & Johnson and C3.ai. Each opens into its architecture, decisions, and outcomes.

    


    

      

        System map · J&J
        
Three production systems, one request path. Point at a node to trace it.

        

          {{ heroCap.tag }}
          {{ heroCap.text }}
        

        

          
            
          
        

      

      
        
        
        
          
          
          
          
          
          
          
        
        
          
          
          
          
          
          
          
        
        
          
          
          
        
        CYCLIC GRAPH
        
          query50K+/day
          routerLangGraph
          agentsA2A
          toolsMCP
          retrievalNeo4j · Qdrant
          evaluationOpenTelemetry
        
      
    


    
    
      
        
  1. 
          
            
            
              {{ p.num }} · {{ p.org }} · {{ p.years }}
              case study
            
            
              {{ p.title }}
              {{ p.problem }}
            
            
              
                {{ m.v }}{{ m.l }}
              
            
            
            
              {{ p.stack }}
              Read case study→
            
          
        
    2. 
          
    

  


  

    

      

        02 // Projects
        
Open source

      

      All 30 repositories on GitHub ↗
    

    

      

        
          
        
      

      
    

    
  
      
No repositories match.

    
    

      {{ pageInfo }}
      
        
      
    

  


  

    

      03 // Experience
      
Where I’ve built

    

    

      
      
      
    
        
          
  1. 
            
            
            
              
    
                {{ e.place }}
                
    
                  
                  
      
                    
                      
    • {{ pt }}
      • 
                          
                  
    
                  
                  
                
    
              
    
                
          
    2. 
            
      

      

        

          
        

        

          Education · 2022–2024
          M.S. Computer Science, University of Southern California
          Graduate Research Assistant: ML analysis of large-scale App Store and developer data for platform acquisitions research. Graduate Teaching Assistant: BUAD 310g Applied Business Statistics.
        

      

    

  


  

    

      

        04 // Certifications
        
Licenses & certifications

        View on LinkedIn ↗
      

      
    
        
          
  • 
            
            
              {{ c.n }}
              {{ c.issuer }}
            
            
              {{ c.date }}
              ↗
            
          
    • 
            
      

    

  


  

    

      05 // Research
      
Publications

    

    

      

        
          
        
        {{ pubCount }}
      

      

        {{ pub.venue }} · {{ pub.year }}
        
{{ pub.title }}

        
{{ pub.body }}

        
          
            
              
              
              
            
            
              
              
              
            
            
              
              
              
              71M+ user reviews100+ semantic keywords per category
              279,000+ developer profilesJaro-Winkler · Levenshtein · 99%
              12,000+ SDKsSDK-to-Category sparse matrices
              
              
              studySMJ 2024
            
          
        
        
          
            
              SITESDEEP LEARNINGDIFFERENTIAL PRIVACYENCRYPTED COMPUTATIONFEDERATED
            
            
            
              
              
              
            
            
              
              
              
            
            
              site 1
              site 2
              site 3
              
              
              
              
              sharedmodel
            
          
        
        {{ pub.note }}
      

    

  


  

    06 // Contact
    

      unnatishah.ix@gmail.com
    

    
    

      © 2026 Unnati Shah · New York City
      {{ palHint }}
    

  








   All work
  

    Case {{ cur.num }} · {{ cur.org }} · {{ cur.years }}
    
{{ cur.title }}

    
{{ cur.problem }}

    

      
        {{ s }}
      
    

  


  

    

      Architecture
    

    
      
        
          
          
          
          
        
        
          
          
          
          
        
        
          
          
          
          
        
        
          
          
          
          orchestratorLangGraph
          intent routing
          agents · A2A
          tools · MCP
          state · cyclic graph
        
      
    
    
      

        

          

            
            
            
          

          
    
            
              
  1. 
                
              
    2. 
                
          

        

      

    
    
      

        
          
        
      

    
    

      {{ cur.stageNum }}
      {{ cur.stageNote }}
    

  


  

    

      My contribution
      

        

          $ unnati.agent run {{ cur.slug }}
          
        

        
    
          
            
  1. {{ c.t }}
    2. 
              
        

        
      

    

    

      Technical decisions
      
    
        
          
  1. {{ dc.n }}{{ dc.t }}
    2. 
            
      

    

    

      Challenges & trade-offs
      
Not covered in the résumé. Add your own notes here: what was hard, what you chose against, and why.

    

    

      Outcomes
      

        
          

            
{{ o.l }}

            
{{ o.v }}

          

        
      

    

    

  

  
    Next case · {{ next.num }}
    {{ next.title }} →
  









  

    $ sudo hire unnati
    [sudo] permission granted. all agents dispatched.
    

      unnatishah.ix@gmail.com
      
        
        
      
    

  




  

    
  




  

    

      
unnati.agent{{ bootPctLabel }}

      
    
        
          
  1. 
            {{ bl.verb }}{{ bl.what }}{{ bl.status }}
          
    2. 
            
      

      

      click or press any key to skip
    

  




  

    

      

        
        
        ESC
      

      
    
        
          
  • 
            {{ it.label }}
            {{ it.kind }} {{ it.mark }}
          
    • 
            
        
  • No matches
    •