Operational complexity, solved with data & AI.
Data engineerAI & automation
Pablo Alejandro López Sánchez
I build forecasting systems, data pipelines and AI agents that run in production.
Manizales, Colombia
12+ stores
Reorder their stock from my forecast every week
Double digit
Percent of forecast error removed from the system it replaced
5 systems
In production, each designed and built end to end
End to end
Ingestion, modelling, orchestration and interface on every one
Selected work
Five systems I designed, built and operate. Four run at Mercaldas, a retail grocery chain in Colombia. The fifth is my own infrastructure. Every one has a working demo you can click.
- 01MLOps · Time series
ML demand forecasting in production
- Series, weekly
- 1,000sSeries, weekly
- % WMAPE cut
- Double digit% WMAPE cut
Replaced the ERP's black box, then caught a silent production failure with an audit I designed myself.
- Python
- LightGBM
- statsforecast
- +5
- 02Data platform · Generative AI
Market price intelligence platform
- Forecast horizon
- 52 wkForecast horizon
- Products covered
- 100sProducts covered
A hostile public data source, turned into a tool the commercial team takes into supplier negotiations.
- Python
- Polars
- Delta Lake
- +5
- 03Applied AI · Guardrails
AI assistant for business data
- Certified + dynamic
- 2 pathsCertified + dynamic
- Figures from the model
- 0Figures from the model
Built the spend guardrails first, then broke them on purpose. The fix moved the control into code.
- Python
- Groq
- Open-weight LLMs
- +2
- 04Backend · Internal tooling
Internal operations platform
- Per plan, was hours
- MinutesPer plan, was hours
- Run by operations
- Self-serveRun by operations
Transfer planning went from a request queued behind me to a tool operations runs on their own.
- Python
- FastAPI
- pandas
- +4
- 05Multi-agent · Local-first
Personal AI operating system
- Required in CI
- 4 gatesRequired in CI
- Cloud spend ceiling
- < $20/moCloud spend ceiling
| 2026-09-04 | Mercado | -186,400 |+ | 2026-09-06 | Energía | -214,300 |approval requiredThe memory is plain text in Git, so it outlives the runtime that reads it.
- Python
- FastAPI
- Next.js
- +5
About
I'm a data and automation engineer at Mercaldas, a retail grocery chain in Colombia, where I build the systems behind forecasting, pricing and internal operations.
Outside work I build my own infrastructure, and I'm interested in how these same methods apply to energy data: load forecasting, distributed generation, and the reliability problems that come with both.
More about how I work