Blog
What we learn building data platforms and putting AI into production on top of them.
- data-mesharchitecturedata-engineering
Data Mesh Without the Complexity: One Bucket Per Team
Implement data mesh with DataSpoc. Each team gets their own bucket, their own pipelines, their own AI agent. No central bottleneck. No platform team required.
2026-04-29 · Michael San Martim
- medalliondata-lakearchitecture
Building a Medallion Architecture Data Lake with DataSpoc
Bronze, Silver, Gold — implement the Databricks medallion pattern using DataSpoc Pipe and Lens. No Spark, no cluster, no cost.
2026-04-29 · Michael San Martim
- ai-agentsagent-mdmcp
The AGENT.md Pattern: Teaching AI Agents to Use Your Tools
How a single Markdown file turns any AI agent into a power user of your data platform.
2026-04-28 · Michael San Martim
- startupdata-lakecost-reduction
The $0 Data Lake for Startups: DataSpoc + S3 in 30 Minutes
A complete data platform for early-stage startups. Ingest from Postgres and Stripe, query with AI, connect agents.
2026-04-28 · Michael San Martim
- airflowetlyaml
10 Lines of YAML vs 200 Lines of Airflow DAG
Side-by-side comparison of building the same ETL pipeline with Airflow and DataSpoc Pipe.
2026-04-28 · Michael San Martim
- multi-clouds3gcs
One Data Lake, Three Clouds: Multi-Cloud Analytics with DataSpoc
Register S3, GCS, and Azure buckets in DataSpoc Lens and query across all of them with one SQL statement.
2026-04-28 · Michael San Martim
- ollamalocal-aiprivacy
Query Your Data Lake with AI for Free: Ollama + DataSpoc Lens
No API key needed. Run AI queries on your data lake using a local LLM with Ollama.
2026-04-28 · Michael San Martim
- windsurfmcpai-agents
Data Analysis from Windsurf IDE with DataSpoc MCP
Connect Windsurf to your data lake. Query, analyze, and build reports without leaving your editor.
2026-04-28 · Michael San Martim
- jupyternotebookduckdb
Jupyter Notebook on Your Data Lake in 60 Seconds
Launch JupyterLab with all your cloud Parquet tables pre-mounted. Zero configuration.
2026-04-27 · Michael San Martim
- incrementalextractionsinger
Incremental Extraction Patterns: Never Re-Extract Old Data Again
How DataSpoc Pipe uses Singer bookmarks to only fetch new data. Save time, bandwidth, and cloud costs.
2026-04-27 · Michael San Martim
- sqltransformsdata-modeling
Building a Curated Data Layer with SQL Transforms (No dbt Required)
Numbered SQL files in DataSpoc Lens replace dbt for simple data transformation workflows.
2026-04-27 · Michael San Martim
- google-sheetsno-codedata-lake
From Google Sheets to a Queryable Data Lake in 3 Commands
Turn any public Google Sheets spreadsheet into Parquet files you can query with SQL.
2026-04-26 · Michael San Martim
- marimonotebookreactive
Reactive Data Analysis with Marimo and DataSpoc Lens
Marimo notebooks + DuckDB on your data lake. Cells update automatically when you change a query.
2026-04-26 · Michael San Martim
- mongodbmigrationparquet
MongoDB to Parquet: Build a Data Lake from Your MongoDB in 15 Minutes
Use DataSpoc Pipe with tap-mongodb to extract collections and convert them to queryable Parquet files.
2026-04-26 · Michael San Martim
- anthropicclaudepython
Building a Data Query Agent with Anthropic Claude SDK and DataSpoc
Use Claude's tool_use feature to let the model query your data lake with real SQL.
2026-04-25 · Michael San Martim
- crewaimulti-agentai-agents
Building a Multi-Agent Data Team with CrewAI and DataSpoc
A DE agent ingests data, a DA agent analyzes it, an ML agent trains models — all powered by DataSpoc.
2026-04-25 · Michael San Martim
- llamaindexai-agentspython
LlamaIndex + DataSpoc: Query Your Data Lake Without Embeddings
Use LlamaIndex's tool-calling agents with DataSpoc Lens SDK for accurate, grounded data queries.
2026-04-25 · Michael San Martim
- duckdbsparkcomparison
DuckDB vs Spark for Data Lake Queries: When Each Wins
DuckDB powers DataSpoc Lens. Spark powers Databricks. Here's when to use each — with benchmarks.
2026-04-24 · Michael San Martim
- mcpclaudeai-agents
How to Build an MCP Server for Your Data Lake
Turn your cloud Parquet files into an API for AI agents with DataSpoc Lens MCP.
2026-04-24 · Michael San Martim
- cursormcpdata-analysis
Analyze Your Data Lake from Cursor IDE with DataSpoc MCP
Connect Cursor to your data lake via MCP. Query tables, explore schemas, and build reports without leaving your editor.
2026-04-23 · Michael San Martim
- duckdbsnowflakecost-reduction
We Replaced Our $8k/month Snowflake with DuckDB and Parquet
How DataSpoc Lens gives you a virtual warehouse over cloud Parquet — at zero cost.
2026-04-23 · Michael San Martim
- fivetranmigrationdata-engineering
Migrating from Fivetran to DataSpoc Pipe: A Step-by-Step Guide
Save $2k/month by replacing Fivetran with open-source DataSpoc Pipe. Same sources, same destinations, zero cost.
2026-04-22 · Michael San Martim
- ragmcpai-agents
RAG vs SQL: Why Your AI Agent Should Query, Not Embed
RAG hallucinates on structured data. Here's why MCP + SQL is more accurate, faster, and cheaper.
2026-04-22 · Michael San Martim
- data-governancesecurityai-agents
Data Governance for AI Agents: How DataSpoc Keeps Your Lake Secure
Read-only access, cloud IAM enforcement, and audit trails — how to let AI agents query without risk.
2026-04-21 · Michael San Martim
- ai-agentslanggraphpython
Building a Data Analyst Agent with LangGraph and DataSpoc Lens
A LangGraph agent that discovers tables, writes SQL, and answers business questions using DataSpoc Lens SDK.
2026-04-21 · Michael San Martim
- data-engineeringetlpostgresql
Postgres to S3 in 5 Minutes with DataSpoc Pipe
Replace your Airflow DAGs with one CLI command. Incremental extraction from PostgreSQL to Parquet in S3.
2026-04-20 · Michael San Martim
- rest-apietlpipeline
Ingest Any REST API to Parquet in 10 Minutes
Use DataSpoc Pipe with tap-rest-api to pull data from any API endpoint into your data lake.
2026-04-20 · Michael San Martim
- autogenmulti-agentanalytics
Multi-Agent Analytics with AutoGen and DataSpoc
Build an AutoGen team where agents collaborate to analyze your data lake — one queries, one visualizes, one reports.
2026-04-19 · Michael San Martim
- claude-codemcpdata-engineering
Using Claude Code as Your Data Engineer: MCP + DataSpoc
Configure Claude Code to ingest data, query your lake, and debug pipelines — all from the terminal.
2026-04-18 · Michael San Martim
- ragparquetduckdb
RAG on Parquet: How to Build Retrieval Over Your Data Lake Without Embeddings
Skip the vector store. Use DataSpoc Lens to give your LLM direct SQL access to Parquet files.
2026-04-17 · Michael San Martim
- langchainsqlpython
LangChain SQLDatabaseChain vs DataSpoc Lens: Which Is Better for Data Queries?
Side-by-side comparison of querying databases with LangChain's SQL chain versus DataSpoc Lens SDK and MCP.
2026-04-16 · Michael San Martim
- openaiai-agentspython
Building a Data Lake Agent with OpenAI Function Calling and DataSpoc
Use OpenAI's function calling to let GPT-4 query your data lake with real SQL via DataSpoc Lens.
2026-04-15 · Michael San Martim