Build in public · Learn by doing · Ship to production

Build an AI Agent Platform
in 10 Phases

Follow along as I build Agent Telar, a public applied AI project. It starts with one Groq API call and grows through RAG, agents and observability. The final phases take it toward a design target of 1 million active users with Kubernetes, Terraform and load testing. Each published phase includes an explanation, interactive flow diagram, quiz and working code.

Start Phase 1 →View on GitHub
4/10 phases live

The Journey

The phases are designed to be completed in order. In the Academy learning path, the next phase unlocks after the previous quiz is passed.

1

AI Gateway

Your first LLM response — streaming, typed, containerized

Node.jsTypeScriptGroq API+2 more
1–2 weeksStart
2
Locked

Smart Agent with Memory

ReAct pattern, tool calling, Redis session memory

LangChain.jsRedisTool Calling+1 more
2–3 weeksComplete AI Gateway first
3
Locked

RAG Knowledge Engine

Document ingestion, pgvector, hybrid search

pgvectorOpenAI EmbeddingsHybrid Search+1 more
3–4 weeksComplete Smart Agent with Memory first
4
Locked

Multi-Agent Orchestration

LangGraph, specialist agents, MCP protocol

LangGraphMCPA2A Protocol
3–4 weeksComplete RAG Knowledge Engine first
5
Coming soon

Production API Platform

Auth, multi-tenancy, usage metering, async queues

JWTBullMQMulti-Tenancy+1 more
2–3 weeks
6
Coming soon

Real-Time Streaming & Queues

Kafka, WebSockets, dead letter queues at scale

KafkaWebSocketsDead Letter Queue+1 more
2–3 weeks
7
Coming soon

Observability & Evaluation

LangSmith, Prometheus, Grafana, LLM-as-judge

LangSmithOpenTelemetryPrometheus+1 more
2–3 weeks
8
Coming soon

Containerized Microservices

Docker, mTLS, service decomposition, CI/CD

DockerGitHub ActionsmTLS+1 more
3–4 weeks
9
Coming soon

Kubernetes Enterprise Deployment

Helm, HPA, KEDA event-driven scaling, ArgoCD GitOps

KubernetesHelmKEDA+1 more
4–6 weeks
10
Coming soon

Terraform IaC + FinOps

AWS infrastructure, cost optimization, unit economics

TerraformAWS EKSSpot Instances+1 more
4–6 weeks

Visual Labs

Alongside the main Agent Telar path, Visual Labs use smaller engineering problems to show how code changes system state, one step at a time.

Explore the TimeService scheduling lab

Built from Real Engineering Work

I am a Senior Full-Stack Engineer and Technical Lead with 10+ years of experience building production web, backend and AWS systems. I work on RAG pipelines and AI deployment on Amazon EKS with AI and infrastructure teams. Telar Academy combines that production background with Agent Telar, which I build and test in public. You can inspect the code, architecture decisions, tests and assumptions behind every published phase.