E. Castellanos
Founder · Strategy lead
Two decades across pricing, forecasting, and ops. Believes a metric you can't name in one sentence isn't a metric.
About
An applied AI lab built by operators, not slide-makers.
ECALabs exists because most AI programs stall in the gap between "interesting result" and "system someone uses on Tuesday." We close that gap. Same team end-to-end, no advisor-to-builder handoff to drop the metric.
What we believe
The shape of useful AI work.
Value 01
Metric before model.
If we can't agree on what we're moving and how we'd measure it, we don't take the work — no exceptions, no "we'll figure it out in flight".
Value 02
Build narrow, learn fast.
A small system in production beats a large one on a roadmap. The first MVP is intentionally smaller than the team would like.
Value 03
Adoption is the deliverable.
Engagement isn't done at deploy. It's done when the workflow runs without us in the room.
Value 04
Direct over decorative.
Plain language, defensible numbers, and one chart that says the thing instead of three that suggest it.
The team
Senior end-to-end. No layered staffing.
Every engagement is staffed with people who have shipped before. We don't backfill juniors learning on your dollar.
M. Aldaña
Head of Engineering
Ships systems, not notebooks. Joined from a quant trading desk where uptime was the only KPI.
P. Werner
Lead Applied Scientist
Forecasting, causal inference, and the quiet work of getting models to behave in production.
L. Okafor
Director, Enablement
Designs adoption programs that read like operating procedures, not seminars.
By the numbers
What "applied AI lab" looks like in practice.
Median MVP
10 weeks
From signed SOW to a system in production with at least one team using it daily.
Engagement shape
Fixed-fee phases
Discovery, build, adoption — each scoped to a single decision worth changing. Renewals happen, never lock-ins.
Team mix
100% senior
No layered staffing. The person on the kickoff is the person on the build review.
Start with a discovery session.
We'd rather scope a focused engagement around a single decision than promise a transformation. The second use case finds itself.