Shift A (06–14)
85.7 kWusual spread ±3.8 kW
Industrial energy analytics
Kinetiqq helps factories find compressed-air waste in the pressure, power and operating logs they already keep. Machine learning learns normal demand across shifts and flags possible leaks and inefficient operation — for engineers to investigate.
Agbara moulding plant · total compressor power · one week
logs only, no sensors added
$11,911 a year — a new leak found in the night and weekend data
Idle-period analysis
At night and at weekends, almost nothing should need compressed air. Whatever the compressors deliver then — beyond a small standby allowance for dryers and controls — is mostly leakage, and it runs every hour of the year.
14 kW of idle demand above standby in a normal week is about $17,872 a year at $0.15/kWh.
standby allowance (8 kW) air used with nothing scheduled
Consumption baselines
Two weeks of logs teach Kinetiqq what each shift and day type normally needs. Later hours are compared with the right baseline — a weekend is not judged against a Monday morning.
Shift A (06–14)
85.7 kWusual spread ±3.8 kW
Shift B (14–22)
80.6 kWusual spread ±4.0 kW
Weeknights
21.9 kWusual spread ±1.1 kW
Weekends
20.8 kWusual spread ±0.9 kW
Header pressure
6.76 barmedian during training
Anomaly detection & investigation priorities
Each alert shows the hours affected, how far demand moved from normal and an estimated yearly cost. Engineers start at the top.
Idle-time demand · Tue 20 Oct 22:00
80 hours · 9.1 kW above normal
Pressure set-point · Tue 27 Oct 06:00 → ongoing
138 hours · 3.0 kW above normal
Outside scheduled hours · Sat 31 Oct 08:00
4 hours · 56.9 kW above normal
Power up, pressure down · Thu 22 Oct 10:00
5 hours · 26.0 kW above normal
Detected in weeks 3–4 of the fictional logs. Costs assume $0.15/kWh; the pressure estimate uses a rule of thumb of about 1% more power per 2 psi.
Repair verification
When maintenance records a repair, Kinetiqq compares idle demand in the week before and the week after — so a leak survey ends with a measured saving, not a guess.
Leak repair · Mon 26 Oct 07:00
−6.9 kW idle demand · ≈ $9,101 a year
How it works
Power, pressure and run-hour exports from the compressor controller or energy meter — CSV is fine.
Baselines are learned per shift and day type from your own weeks of data.
Idle-time demand, production excess, unscheduled running and pressure drift, ranked by cost.
Engineers confirm or dismiss each alert; repairs are checked against the data afterwards.
Every alert is a reason to look, confirmed or dismissed by an engineer.
Baseline, measured demand, hours and cost method are shown for every alert.
The first version uses logs you already have; meters and sensors can come later.
Written summaries come only from the calculated numbers.
Why it matters
Share of a compressor’s output that leaks can sometimes waste.
Compressed Air Challenge, Fact Sheet 7: Compressed Air System Leaks (2026)Typical leak rate of a plant that has not been well maintained, as a share of total compressed-air production capacity.
Compressed Air Challenge, Fact Sheet 7 (2026)Leakage that proactive leak detection and repair can bring a system down to.
Compressed Air Challenge, Fact Sheet 7 (2026)Who it’s for
Plants with compressors running around the clock and logs nobody has time to read.
Audits that start from data before anyone walks the site.
Leak surveys aimed where the data points — and savings you can prove afterwards.
Pricing
Illustrative pricing for a proposed product, to be agreed with pilot customers.
$350per site / month
$990per month
Custom
Technology
Kinetiqq works on time series that never stop: compressors log power and pressure every few minutes, around the clock, at every site. Designing on AWS lets log storage, anomaly models, scheduled analysis and model access run as managed services, and keeps each site’s data isolated as customers grow from one plant to a portfolio of sites.
Amazon S3
Compressor power, pressure and operating-schedule exports land in per-site storage, partitioned by date.
AWS Lambda
Lambda runs on a schedule to validate new logs, align them with production calendars and compute idle-period demand.
Amazon SageMaker AI
Models learn normal demand per shift and day type, then score new periods for unusual power and pressure patterns.
Amazon DynamoDB
Compressors, set-points, alerts, engineer decisions and repair records are stored per site.
Amazon Bedrock
Plain-language summaries of each alert are drafted from the calculated numbers for engineers to review.
Across every stage
Proposed, not deployed. This website and the air audit run in your browser on fictional data.
Team
Founder
Tifinie founded Kinetiqq after seeing plants pay to run compressors all weekend for air nobody was using — while the logs that showed it sat unread. Tifinie leads product, customers and every pilot.
Progress & pilot programme
Kinetiqq is in its building stage and opening its first pilots. A pilot covers one site: you upload a few weeks of compressor logs, we learn the baselines, and your engineers review every alert — then we verify the repairs together.
By Tifinie Taylors, to find compressed-air waste in data plants already collect.
Shift-by-shift baselines, idle-period analysis, pressure checks and cost-ranked alerts.
Alerts, engineer review and repair verification shown on four weeks of fictional logs.
Log intake, anomaly models, equipment records, scheduled analysis and evidence summaries designed.
One site per pilot: uploaded compressor logs, with every alert reviewed by your engineers.
FAQ
Compressor power (or current) and header pressure logs, ideally at 15-minute intervals or better, plus your production schedule. A few weeks is enough to start.
Not for the first version. Kinetiqq works from logs your controllers or energy meters already record.
No. It tells you when demand changed and how much it costs; an ultrasonic survey finds the leak. Kinetiqq then checks the repair worked.
Not yet. It is a proposed product; the air audit is a prototype running in your browser on fictional logs.
Contact
Questions about the product, the pilot programme, log formats or how operating data is handled. Messages go directly to the founder.