JOBWAVE VERIFIED

Applied AI Engineer II

Company: Deloitte
Requisition Code: 365121
Team: US Deloitte Technology Product Engineering
Type: Full Time
Experience: 5+ Years
GenAI / Agentic Experience: 3+ Years
Travel: ~10% (business/product dependent)
Immigration Sponsorship: Limited availability
Location: Not specified
Salary: Not specified

────────────────────

JOBWAVE INSIGHTS

Who is this for?

An experienced full-stack software engineer with 5+ years of engineering experience and 3+ years building production Generative AI and Agentic AI systems.

The role is designed for someone who can move beyond experimentation and build production-grade AI products, combining full-stack engineering, LLMs, RAG, multi-agent systems, cloud architecture, observability, evaluation, security, and FinOps.

Key Skills

• Python
• Node.js
• C# / .NET
• Java
• React
• Angular
• SQL
• NoSQL
• REST APIs
• GraphQL
• Microservices
• Object-Oriented Design
• Data Structures & Algorithms
• Generative AI
• LLMs
• Agentic AI
• LangChain
• LangGraph
• RAG
• Vector Databases
• Multi-Agent Systems
• Tool Calling
• Human-in-the-Loop
• LLM Evaluation
• LangSmith
• LangFuse
• MLflow
• PyTorch
• TensorFlow
• AWS
• Azure
• GCP
• Azure OpenAI
• AWS Bedrock
• Vertex AI
• Infrastructure as Code
• FinOps
• Token Optimization
• CI/CD
• DevSecOps
• SRE
• GitHub
• Azure DevOps
• SonarQube

Priority Skills

Highest priority:

• Production Generative AI
• Agentic AI / Multi-Agent Systems
• Python or another major backend language
• Full-stack development
• RAG
• Vector databases
• LangChain / LangGraph
• LLM evaluation
• Cloud-native engineering
• Microservices
• System design
• AI observability
• Token and inference-cost optimization

Strong advantages:

• LangSmith / LangFuse
• MLflow
• PyTorch / TensorFlow
• Azure OpenAI
• AWS Bedrock
• Vertex AI
• Infrastructure as Code
• DevSecOps
• SRE
• Semantic caching
• Hybrid search
• Re-ranking
• Quantized / smaller models
• LLM output testing

What to Prepare

• Full-stack system design
• Object-Oriented Design
• Data Structures & Algorithms
• Microservice architecture
• REST / GraphQL APIs
• SQL and NoSQL databases
• RAG architecture
• Vector search
• Hybrid search
• Chunking strategies
• Re-ranking
• Multi-agent orchestration
• LangGraph
• LangChain
• Tool calling
• Agent state management
• Human-in-the-loop workflows
• LLM evaluation
• AI observability
• Token optimization
• Inference latency optimization
• Semantic caching
• Model selection
• Cloud AI platforms
• AWS Bedrock
• Azure OpenAI
• Vertex AI
• CI/CD
• DevSecOps
• SRE
• Automated testing

What to Focus On

1. Production AI Engineering

The role is focused on production AI systems, not just prompt engineering.

Be prepared to explain the complete lifecycle:

User → Application → Agent → LLM → Tools / APIs → RAG → Data → Response → Evaluation → Monitoring

2. RAG Architecture

Understand the full RAG pipeline:

• Document ingestion
• Chunking
• Embedding
• Vector storage
• Retrieval
• Hybrid search
• Re-ranking
• Context construction
• LLM generation
• Evaluation

Be prepared to discuss how you reduce hallucinations and improve retrieval quality.

3. Multi-Agent Systems

Understand:

• Agent orchestration
• Agent roles
• State management
• Tool calling
• Agent-to-agent communication
• Failure handling
• Human-in-the-loop escalation
• Observability
• Evaluation

Practice building a small multi-agent application using LangGraph or LangChain.

4. Full-Stack Engineering

Strong AI knowledge is not enough for this role.

Prepare for:

• Backend API design
• Microservices
• Frontend integration
• SQL / NoSQL databases
• Vector databases
• Authentication and authorization
• API scalability
• Error handling
• System design

5. FinOps & Token Economics

Understand how AI applications create costs through:

• Input tokens
• Output tokens
• Model selection
• Number of LLM calls
• Retrieval operations
• Agent loops
• Inference latency
• Cloud infrastructure

Prepare optimization strategies such as:

• Smaller models
• Quantized models
• Semantic caching
• Prompt optimization
• Context reduction
• Request batching
• Streaming
• Model routing

6. Evals & Observability

Know how to measure AI system quality.

Prepare for:

• LLM evaluation
• Retrieval evaluation
• Hallucination detection
• Response quality
• Agent traces
• Latency monitoring
• Token usage
• Error monitoring

Understand tools such as:

• LangSmith
• LangFuse
• MLflow

7. Cloud Architecture

Be prepared to compare:

• AWS vs Azure vs GCP
• AWS Bedrock
• Azure OpenAI
• Vertex AI

Understand how to deploy AI applications using cloud-native architecture.

8. DevSecOps & Engineering Quality

Prepare to discuss:

• CI/CD
• Automated testing
• Security
• Infrastructure as Code
• Code quality
• SRE
• Monitoring
• Vulnerability management
• AI-specific testing

Potential Gaps

Candidates may need additional preparation if they have:

• Strong software engineering but limited production GenAI experience
• Prompt-engineering experience without production AI architecture
• Limited multi-agent experience
• Weak RAG knowledge
• No vector database experience
• Limited LLM evaluation experience
• No AI observability experience
• Weak cloud architecture knowledge
• Limited FinOps/token optimization experience
• Limited microservice architecture experience
• Weak system design fundamentals
• No DevSecOps experience
• Limited automated LLM testing
• Little experience with LangChain or LangGraph

Key experience requirement to verify: 5+ years of full-stack software engineering and 3+ years specifically focused on production GenAI/Agentic systems.

────────────────────

JOB DETAILS

Original Role Information

Role: Applied AI Engineer II
Requisition Code: 365121
Team: US Deloitte Technology Product Engineering
Travel: ~10% depending on business/product
Immigration Sponsorship: Limited availability

The role focuses on building production-grade full-stack products that integrate Generative AI, Agentic Workflows, RAG, multi-agent systems, and cloud-native technologies.

Responsibilities

Full-Stack & AI Architecture

• Design full-stack web platforms.
• Build backend microservices.
• Integrate LLMs into production applications.
• Build RAG pipelines.
• Develop multi-agent systems.
• Deploy production AI solutions.

FinOps & Optimization

• Manage cloud infrastructure efficiency.
• Optimize inference latency.
• Reduce token consumption.
• Monitor AI operational costs.
• Balance technical performance with cost-to-value.
• Optimize model and infrastructure selection.

Agentic SSDLC & DevSecOps

• Implement CI/CD pipelines.
• Apply spec-driven development.
• Implement DevSecOps practices.
• Use AI-assisted development tools.
• Integrate automated testing into the development lifecycle.

Rapid Prototyping

• Build lightweight prototypes.
• Validate concepts quickly.
• Gather early user feedback.
• Iterate rapidly.
• Avoid unnecessary upfront engineering when validating new concepts.

Cross-Functional Collaboration

• Work with UX designers.
• Collaborate with product managers.
• Partner with business stakeholders.
• Balance technical feasibility with user experience.
• Align technical solutions with product strategy.

────────────────────

REQUIREMENTS

Experience

• 5+ years of full-stack software engineering experience.
• 3+ years building production Generative AI / Agentic AI systems.
• 3+ years of cloud-native engineering experience.
• Practical production experience rather than purely academic AI knowledge.

Education

Bachelor's degree in:

• Computer Science
• Software Engineering
• Data Science
• Machine Learning
• Related quantitative field

Full-Stack Engineering

Experience with one or more:

• Python
• Node.js
• C# / .NET
• Java
• React
• Angular

Also required:

• SQL databases
• NoSQL databases
• OOP / OOD
• Data Structures & Algorithms
• UML
• Software architecture

Generative AI & Agentic Systems

• Production LLM applications
• OpenAI
• Anthropic
• Open-source LLMs
• LangChain
• LangGraph
• PyTorch
• TensorFlow
• RAG
• Vector databases
• Multi-agent orchestration
• Model evaluations
• AI observability
• LangSmith
• LangFuse

Cloud & FinOps

• AWS, Azure, or GCP
• Azure OpenAI
• AWS Bedrock
• Vertex AI
• Cloud-native architecture
• Infrastructure as Code
• Token optimization
• Latency optimization
• Cloud cost management

Engineering Tooling

• GitHub
• Azure DevOps
• SonarQube
• MLflow
• CI/CD
• DevSecOps
• SRE
• Lean / XP Agile

────────────────────

QUALIFICATIONS

• Bachelor's degree in a relevant technical or quantitative discipline.
• 5+ years full-stack software engineering experience.
• 3+ years production GenAI/Agentic AI experience.
• 3+ years cloud-native engineering experience.
• Strong software architecture fundamentals.
• Experience developing production LLM applications.
• Experience with RAG and vector databases.
• Experience with multi-agent systems.
• Experience with AI evaluation and observability.
• Understanding of FinOps and AI cost optimization.
• Experience with CI/CD and DevSecOps.
• Strong cross-functional collaboration skills.

────────────────────

INTERVIEW PREPARATION

Phase 1 — Full-Stack & System Design

Revise:

• OOP/OOD
• DSA
• Clean architecture
• Microservices
• REST / GraphQL
• SQL / NoSQL
• ER diagrams
• Sequence diagrams
• Component diagrams

Practice designing an AI-powered application from frontend to database and LLM layer.

Phase 2 — RAG & Multi-Agent Systems

Build a small project using:

React → API → LangGraph/LangChain → RAG → Vector DB → LLM

Include:

• Multiple agents
• State management
• Tool calling
• Human approval
• Error handling
• Retrieval
• Evaluation

Phase 3 — FinOps, Evals & Observability

Prepare to explain:

• How you reduce token consumption
• How you reduce LLM latency
• How you select models based on cost and quality
• How semantic caching works
• How smaller/quantized models can reduce cost
• How streaming improves perceived latency
• How you evaluate LLM outputs
• How you trace agent behavior

Practice with LangSmith, LangFuse, or MLflow.

Phase 4 — Architecture & Case Study

Prepare to defend trade-offs such as:

• AWS Bedrock vs Azure OpenAI
• Proprietary vs open-source models
• RAG vs fine-tuning
• Single-agent vs multi-agent
• Synchronous vs asynchronous processing
• Large vs smaller models
• SQL vs NoSQL
• Vector DB choices
• Serverless vs containerized deployment

For each decision, explain:

Requirement → Options → Trade-offs → Decision → Cost → Performance → Security → Business Impact