Babasola Osibo has spent more than two decades working with high-stakes infrastructure, including telecom networks, data centers, and the systems that keep them running. His specialty is simple to describe and hard to achieve: using data and machine learning to make critical environments faster, steadier, and more power-efficient.
Osibo’s peer-reviewed paper, “Transforming High-Energy Data Center Sites: Sustainability with Predictive Analytics and Futuristic Technologies,” outlines a practical framework that any facility can adopt. No exotic hardware. No moonshots. Just disciplined telemetry, compact ML models, and controls with guardrails backed by verification you can audit.
Why this matters now
Compute demand is climbing, thanks to AI, cloud services, and an always-on digital life. For operators, that usually means higher energy bills and tougher reliability targets. The question is how to bend the power curve without gambling on service levels.
Osibo’s answer is to stop reacting and start anticipating. His framework focuses on three lever operations teams already touch every day and shows how to wire each one to measurable, analytics-driven savings.
Move 1: Forecast Demand Before It Bites
Short-horizon models predict near-term server load, allowing teams to stage capacity and cooling in advance. This reduces unnecessary overprovisioning, smooths peaks, and protects SLAs when traffic surges.
How it works: minute-level IT load and seasonality features feed lightweight time-series and gradient-boosted models to generate 60-minute forecasts with conservative error bands. Inputs include maintenance flags, as well as outside-air and wet-bulb readings. The output is a clear, actionable forecast that operations can trust.
Move 2: Place Workloads Where They Cost Less
Not all capacity is equal. By routing jobs to lower-cost or lower-carbon regions, within strict reliability envelopes, operators can trim kilowatt-hours without touching application performance.
How it works: a data-driven placement policy pairs the demand forecast with queue depth, latency budgets, and regional price/carbon signals, then selects targets with thermal headroom, ensuring that latency and failover rules are never violated. The result is smoother curves and fewer expensive spikes.
Move 3: Tune Cooling Like a Control System
Cooling can be a third to nearly half of a data center’s energy use. Osibo recommends set point adjustments guided by live signals, including rack-inlet temperatures, chiller efficiency (COP), airflow behavior, and outside conditions.
How it works: response models (regression/GBM) estimate the kWh impact of a ±0.5–1.0 °C setpoint nudge under current weather and load. Changes are bounded by standard thermal envelopes and auto-rollback if sensors drift. Small, reversible moves add up without ever leaving safe operating ranges.
How the System Fits Together
Osibo describes a pipeline that operators will recognize:
Telemetry → Features → Forecast → Constrained Control → Verify.
Minute-level signals, IT load, rack temperatures, CRAC/CRAH setpoints, chiller COP, airflow hints, outside-air/wet-bulb, grid price/renewables, feed compact models that propose changes. Those proposals are bound by widely used thermal limits and site SLAs. Weekly retraining and drift checks keep models honest. Human-in-the-loop approvals and clean rollback plans keep operations reversible. It’s an operations-first pattern designed for trust.
Proof You Can Stand Up in a Meeting
Before moving to the United States, Osibo spent years building and running various practices at scale. At MTN, one of Africa’s largest network operators, he progressed from frontline engineering to regional leadership for enterprise services. The internal dashboards he oversaw reported approximately 20% less network downtime, roughly 25% faster deployments (SLA compliance rising from around 80% to over 98%), and approximately 25% lower electricity use across 30-plus data centers and around 2,000 base-station sites. He also trained over 2,500 engineers across West Africa, transforming new methods into everyday routines.
Those results weren’t accidents. They came from standardizing the data path (signals → features), the model cadence (retrain/monitor), the approval gates, and the measurement plan, so finance and sustainability teams can audit the savings. In other words: make efficiency an engineering discipline.
The Pedigree Behind the Framework
Osibo’s path blends hands-on operations with formal analytics. He earned a B.Sc. in Agricultural Engineering (University of Ibadan), a Master of Business Leadership (University of South Africa), and an M.S. in Business Analytics (University of Dallas). He holds a PMP, CSM, CCNA, ITIL, and Lean Six Sigma Green Belt credentials; is a Senior Member of IEEE; and is a Fellow of the Institute of Management Consultants. His book, Python Essentials: A Practical Guide, is cataloged by the McKinney Public Library System.
What Good Looks Like (and How to Prove It)
Osibo’s paper emphasizes verification as much as technique. Select metrics that operators already track, including energy (PUE/DCiE), carbon (CUE), water (WUE), alongside reliability metrics (MTTR, SLA). Establish a weather-normalized baseline. Pilot changes on a small slice of capacity with a like-for-like control. Publish results with a simple confidence range. The goal is not a flashy number; it’s a number that survives scrutiny.
“You earn trust by being boring in the right places,” Osibo says. “Clear limits, safe defaults, and runbooks everyone can follow.”
The Bigger Picture
As data centers grow, credible efficiency is no longer optional; it’s table stakes. Investors want durable savings, regulators want transparency, and customers want reliability that doesn’t spike the bill. Babasola Osibo’s contribution is a common language and a set of analytics-powered moves teams can adopt with the tools they already have.
Predict, prevent, protect. That’s the through-line in his work, from on-call duties to publishing a framework others can reuse. For operators deciding what to do next quarter, the roadmap is straightforward: build the analytics, constrain the controls, and verify the gains.
