← All paths

Path 03 · full reference map

Production AI & Data Platforms

Applied RAG, agents, evaluation, Spark, Delta Lake, MLflow, and data systems assembled into production-grade architectures.

7 modules56 topicsSelf-paced90-day applied AI systems guide

Prior practice · disclosed carefully

I am not starting this path at zero.

These claims come from résumé-backed work. They are intentionally generalized, so the site shows what I have applied without inventing public artifacts or exposing private systems.

EvalOps

Building an open-source evaluation-operations platform around Delta Lake, MLflow, RAGAS, and Langfuse.

Data pipelines

Built streaming and analytical pipelines using Apache Beam, ClickHouse, PostgreSQL, and event-driven infrastructure.

Applied AI

Integrated multimodal AI into a production workflow, cutting a manual operational process from minutes to near-real-time.

Published progress

18%

10 of 56 topics complete · updated with each site release

01

Weeks 01-02

RAG architecture

Build retrieval end to end and compare the choices empirically.

1/8
Document ingestionNot yet published
Fixed and semantic chunkingNot yet published
Embedding selectionNot yet published
Chroma and pgvectorNot yet published
Hybrid retrievalNot yet published
RerankingNot yet published
Streaming generationNot yet published
RAGAS baselinePracticed · complete
02

Week 03

Spark and Delta Lake

Add the data-engineering substrate behind AI applications.

0/8
Spark execution modelNot yet published
DataFrames and Spark SQLNot yet published
PySpark transformationsNot yet published
Delta tablesNot yet published
Schema enforcementNot yet published
Time travelNot yet published
Bronze/silver/gold layersNot yet published
Structured StreamingNot yet published
03

Week 04

Stateful agents

Build tool-using workflows with explicit state and failure handling.

0/8
LangChain primitivesNot yet published
LangGraph stateNot yet published
Tool contractsNot yet published
Persistent memoryNot yet published
Redis stateNot yet published
Fallback pathsNot yet published
Human escalationNot yet published
Trace instrumentationNot yet published
04

Week 05

Evaluation platform

Generalize evaluation into reusable infrastructure.

6/8
RAGASPracticed · complete
Custom quality metricsNot yet published
Agent trace evaluationNot yet published
Prompt-version regressionPracticed · complete
Cost trackingPracticed · complete
MLflow experimentsPracticed · complete
Langfuse tracesPracticed · complete
Evaluation CLIPracticed · complete
05

Week 06

Multi-agent and MCP

Study orchestration, specialization, and protocol boundaries.

0/8
Supervisor-worker topologyNot yet published
Specialist routingNot yet published
Structured messagesNot yet published
MCP serversNot yet published
Tool authorizationNot yet published
Cost-aware routingNot yet published
Loop detectionNot yet published
Multi-agent evaluationNot yet published
06

Weeks 07-08

Latency and cost

Optimize the application path with evidence.

0/8
Semantic cachingNot yet published
Model routingNot yet published
Prompt compressionNot yet published
BatchingNot yet published
Token budgetsNot yet published
Latency percentilesNot yet published
DSPy optimizationNot yet published
Baseline comparisonsNot yet published
07

Weeks 09-12

Capstone platform

Join data, retrieval, evaluation, and operations in one system.

3/8
Architecture documentNot yet published
API specificationNot yet published
Data pipelinePracticed · complete
Retrieval serviceNot yet published
Agent orchestrationNot yet published
Evaluation servicePracticed · complete
Observability dashboardPracticed · complete
Deployment and runbookNot yet published