Engineering Agents
Frontend, Backend, Database, AI, DevOps, and Firmware engineers
🎯 Key Opinion: These agents work best when given specific, scoped tasks with clear acceptance criteria. Vague requests produce vague results. Take the time to define what "done" looks like before invoking an engineering agent.
Overview
Engineering agents are the implementation layer of your AI team. Each agent specializes in a specific domain, bringing deep knowledge of best practices, common patterns, and technology-specific idioms. They turn technical specifications into working code.
| Agent | Domain | Technologies |
|---|---|---|
| Frontend Engineer | UI/UX, components, state | React, Vue, Svelte, Next.js, CSS |
| Backend Engineer | APIs, services, logic | Node.js, Python, Go, Java, Rust |
| Database Engineer | Data modeling, queries | PostgreSQL, MongoDB, Redis, MySQL |
| AI Engineer | ML, LLMs, RAG | OpenAI, Anthropic, LangChain, Pinecone |
| DevOps Engineer | Infrastructure, CI/CD | Docker, Kubernetes, Terraform, GitHub Actions |
| Firmware Engineer | Embedded systems | C, C++, Rust, ARM, RTOS |
Frontend Engineer
The Frontend Engineer specializes in user interfaces, from component architecture to accessibility and performance optimization. It understands modern frontend frameworks and CSS methodologies.
Responsibilities
| Area | Capabilities |
|---|---|
| Components | Reusable UI components, design systems, atomic design patterns |
| State Management | Redux, Zustand, Jotai, React Query, context patterns |
| Routing | Client-side routing, nested routes, protected routes, SSR/SSG |
| Accessibility | ARIA, keyboard navigation, screen readers, WCAG compliance |
| Performance | Code splitting, lazy loading, bundle optimization, Core Web Vitals |
Technologies
- Frameworks: React, Vue.js, Svelte, Angular, Solid.js
- Meta-frameworks: Next.js, Nuxt, SvelteKit, Remix, Astro
- Styling: Tailwind CSS, CSS Modules, Styled Components, Sass
- Build Tools: Vite, Webpack, esbuild, Turbopack
- Testing: Jest, Vitest, Testing Library, Playwright
Invoking the Frontend Engineer
# Build a new component
@frontend-engineer Create a DataTable component with:
- Sortable columns
- Pagination (10/25/50 per page)
- Row selection with checkboxes
- Loading and empty states
Use React, TypeScript, and Tailwind CSS.
# Optimize performance
@frontend-engineer Audit and optimize the dashboard page.
Current LCP is 4.2s, target is under 2.5s.
Focus on code splitting and image optimization.
# Implement accessibility
@frontend-engineer Make the checkout form WCAG 2.1 AA compliant.
Add proper ARIA labels, keyboard navigation, and error announcements.📋 Best Practice: Include the tech stack in your request. The Frontend Engineer adapts its output to your specific framework and styling approach.
Backend Engineer
The Backend Engineer handles server-side logic, API design, integrations, and authentication systems. It builds scalable, secure services that power your application.
Responsibilities
| Area | Capabilities |
|---|---|
| APIs | REST, GraphQL, gRPC, WebSocket, API versioning, rate limiting |
| Business Logic | Domain modeling, validation, workflows, state machines |
| Integrations | Third-party APIs, webhooks, message queues, event buses |
| Authentication | JWT, OAuth 2.0, OIDC, session management, RBAC, ABAC |
| Security | Input validation, SQL injection prevention, CORS, CSP, secrets management |
Technologies
- Node.js: Express, Fastify, NestJS, Hono
- Python: FastAPI, Django, Flask, Starlette
- Go: Gin, Echo, Fiber, Chi
- Java: Spring Boot, Quarkus, Micronaut
- Rust: Actix, Axum, Rocket
Invoking the Backend Engineer
# Create an API endpoint
@backend-engineer Create a POST /api/orders endpoint that:
- Validates order items and quantities
- Calculates totals with tax
- Creates order record in database
- Publishes OrderCreated event
- Returns order confirmation with ID
Use TypeScript, Express, and Zod for validation.
# Implement authentication
@backend-engineer Add JWT authentication with:
- Login endpoint with email/password
- Refresh token rotation
- Password reset flow
- Rate limiting on auth endpoints
# Build an integration
@backend-engineer Integrate with Stripe for payment processing.
Support: card payments, refunds, webhooks for async events.⚠️ Security Note: Always specify security requirements explicitly. The Backend Engineer implements secure defaults, but edge cases need explicit guidance.
Database Engineer
The Database Engineer designs schemas, writes efficient queries, manages migrations, and optimizes database performance. It understands both relational and NoSQL paradigms.
Responsibilities
| Area | Capabilities |
|---|---|
| Schema Design | Normalization, denormalization, relationships, constraints, indexes |
| Query Optimization | Query plans, index selection, N+1 prevention, batch operations |
| Migrations | Schema versioning, zero-downtime migrations, rollback strategies |
| Performance | Indexing strategies, partitioning, connection pooling, caching |
Technologies
- Relational: PostgreSQL, MySQL, SQLite, SQL Server
- NoSQL: MongoDB, DynamoDB, Cassandra, CouchDB
- Cache/KV: Redis, Memcached, KeyDB
- Search: Elasticsearch, Meilisearch, Algolia
- ORMs: Prisma, Drizzle, TypeORM, SQLAlchemy, GORM
Invoking the Database Engineer
# Design a schema
@database-engineer Design a schema for a multi-tenant SaaS app.
Entities: organizations, users, projects, tasks.
Requirements: soft deletes, audit logging, row-level security.
Use PostgreSQL with Prisma.
# Optimize a slow query
@database-engineer This query takes 3s on 1M rows:
SELECT * FROM orders
JOIN order_items ON orders.id = order_items.order_id
WHERE orders.created_at > NOW() - INTERVAL '30 days'
Analyze and optimize to under 100ms.
# Create a migration
@database-engineer Create a zero-downtime migration to add
a 'status' column to the orders table with default 'pending'.
Include rollback script.💡 Pro Tip: Share table sizes and current query plans when asking for optimization help. Context about data volume dramatically improves recommendations.
AI Engineer
The AI Engineer builds intelligent features using LLMs, RAG systems, and machine learning. It handles prompt engineering, model selection, and evaluation pipelines.
Responsibilities
| Area | Capabilities |
|---|---|
| LLM Integration | API integration, streaming, function calling, structured outputs |
| RAG Systems | Vector stores, embeddings, chunking strategies, retrieval tuning |
| Prompt Engineering | System prompts, few-shot examples, chain-of-thought, prompt templates |
| Evaluation | Metrics design, A/B testing, regression detection, human evaluation |
Technologies
- LLM Providers: OpenAI, Anthropic, Google, Mistral, Cohere
- Frameworks: LangChain, LlamaIndex, Semantic Kernel, Haystack
- Vector DBs: Pinecone, Weaviate, Qdrant, Chroma, pgvector
- Embeddings: OpenAI, Cohere, Voyage AI, sentence-transformers
- Evaluation: LangSmith, Braintrust, Promptfoo
Invoking the AI Engineer
# Build a RAG system
@ai-engineer Build a documentation Q&A system:
- Ingest markdown files from docs/
- Use OpenAI embeddings with Pinecone
- Implement semantic search with reranking
- Return answers with source citations
# Create a prompt template
@ai-engineer Design a prompt for extracting structured data
from customer support emails. Extract: sentiment, topic,
urgency, action items. Output as JSON.
# Set up evaluation
@ai-engineer Create an evaluation pipeline for our chatbot.
Metrics: relevance, groundedness, helpfulness.
Include automated scoring and human review workflow.📋 Best Practice: Always include expected input/output examples when requesting prompt engineering. The AI Engineer uses these for few-shot learning and testing.
DevOps Engineer
The DevOps Engineer manages infrastructure, CI/CD pipelines, containerization, and monitoring. It builds reliable, scalable deployment systems.
Responsibilities
| Area | Capabilities |
|---|---|
| CI/CD | Build pipelines, test automation, deployment strategies, rollbacks |
| Infrastructure as Code | Terraform, CloudFormation, Pulumi, resource provisioning |
| Containers | Docker, Kubernetes, ECS, service mesh, scaling policies |
| Monitoring | Metrics, logging, tracing, alerting, SLOs/SLAs |
Technologies
- CI/CD: GitHub Actions, GitLab CI, Jenkins, CircleCI
- IaC: Terraform, Pulumi, AWS CDK, CloudFormation
- Containers: Docker, Kubernetes, Helm, ArgoCD
- Cloud: AWS, GCP, Azure, Vercel, Fly.io
- Monitoring: Datadog, Grafana, Prometheus, PagerDuty
Invoking the DevOps Engineer
# Create a CI/CD pipeline
@devops-engineer Create a GitHub Actions workflow for:
- Lint and test on PR
- Build and push Docker image on merge to main
- Deploy to staging automatically
- Manual promotion to production with approval
# Set up infrastructure
@devops-engineer Create Terraform for a production setup:
- ECS Fargate cluster
- RDS PostgreSQL with read replicas
- ElastiCache Redis
- ALB with SSL termination
Use modules and follow AWS best practices.
# Configure monitoring
@devops-engineer Set up observability stack:
- Application metrics with Prometheus
- Log aggregation with Loki
- Distributed tracing with Jaeger
- Dashboards in Grafana with alerts⚠️ Important: Always review infrastructure changes before applying. Use
terraform plan output to verify changes match expectations.Firmware Engineer
The Firmware Engineer develops embedded systems software, from hardware interfaces to real-time operating systems. It handles the unique constraints of resource-limited environments.
Responsibilities
| Area | Capabilities |
|---|---|
| Hardware Interfaces | GPIO, I2C, SPI, UART, ADC/DAC, PWM, DMA |
| RTOS | FreeRTOS, Zephyr, ThreadX, task scheduling, synchronization |
| Drivers | Sensor drivers, display drivers, communication peripherals |
| Protocols | BLE, WiFi, LoRa, CAN, Modbus, MQTT, CoAP |
Technologies
- Languages: C, C++, Rust, Assembly
- Platforms: ARM Cortex-M, ESP32, STM32, Nordic nRF, Raspberry Pi Pico
- RTOS: FreeRTOS, Zephyr, NuttX, ThreadX
- Tools: GCC, LLVM, OpenOCD, J-Link, Logic Analyzers
- Frameworks: Arduino, PlatformIO, ESP-IDF, Mbed
Invoking the Firmware Engineer
# Create a driver
@firmware-engineer Write a driver for BME280 sensor over I2C.
Platform: STM32F4, using HAL.
Features: temperature, humidity, pressure readings.
Include power management (sleep/wake).
# Implement a protocol
@firmware-engineer Implement BLE GATT server for a fitness tracker.
Services: heart rate, step count, battery level.
Platform: Nordic nRF52840 with Zephyr RTOS.
# Optimize for power
@firmware-engineer Optimize this firmware for battery life.
Current: 15mA average. Target: under 5mA.
MCU: ESP32-S3. Application: environmental sensor logging.📋 Hardware Context: Always specify the exact MCU, peripherals, and constraints. Firmware is highly hardware-specific—generic advice rarely applies directly.
Coordination Patterns
Engineering agents work best when coordinated through the Work Orchestrator. Here are common patterns for multi-agent tasks:
Full-Stack Feature
# Parallel implementation of a complete feature
@work-orchestrator Coordinate implementation of user profile feature:
Phase 1 (parallel):
@database-engineer: Create users table with profile fields
@frontend-engineer: Build profile form component skeleton
Phase 2 (parallel, after Phase 1):
@backend-engineer: Create CRUD endpoints for profiles
@frontend-engineer: Implement form validation and state
Phase 3 (after Phase 2):
@frontend-engineer: Integrate form with API
@qa-backend: Write API integration tests
@qa-frontend: Write E2E testsAI-Powered Feature
# Adding AI capabilities to existing system
@work-orchestrator Coordinate AI search implementation:
Phase 1:
@ai-engineer: Set up vector store and embedding pipeline
@database-engineer: Add vector column to documents table
Phase 2 (after Phase 1):
@backend-engineer: Create search API endpoint
@ai-engineer: Implement semantic search with reranking
Phase 3 (after Phase 2):
@frontend-engineer: Build search UI with result highlighting