MachineMary
AI Troubleshooting Copilot for the Shop Floor

25 downtime events a month.
Your operators handle zero
on their own.

Machine Mary guides operators through troubleshooting — step by step, grounded in your machines and your fault history. Every event becomes a record. Knowledge compounds. It doesn't retire, forget, or vary by shift.

Currently piloting with leading auto-component manufacturers.

Book a 30-min demo→ See how it works↓
7 min
Avg. resolution with Mary
40–60 min
Without
24/7
Every shift, every line
Live · 02:14 AM Voice or text input
Tap a step to flip through the prototype
Before vs. After · same machine, two shifts

Two shifts. Same machine. One has Mary.

The dramatic save matters. The routine save matters more — that's where the volume lives.

×
Before
Without Mary
VS
✓
After
With Mary
Without Mary 2:14 AM · Tue
The 2 AM phone call.
Operator hears an unusual noise from Station 12. Tries restarting — no luck. Line leader takes a look — not sure either. They call maintenance at home. Tech drives in, asks "what happened?" Operator re-explains. Tech re-diagnoses from scratch, finds a material jam in the feed assembly. Clears it in 3 minutes.
Downtime · cumulative00:00 → 00:60
40–60min · no record
With Mary 2:14 AM · Tue
No phone call. 7 minutes.
Operator opens Mary, taps Station 12, records: "weird noise, machine stopped." Mary matches the pattern — two options: (A) check feed assembly for jam, (B) check drive belt tension. Operator taps A, clears the jam, line restarts.
Downtime · cumulative00:00 → 00:07
7min · fault logged
Without Mary 10:47 AM · Tue · routine
The error code nobody remembers.
Operator sees an error code on the HMI. Not sure what it means. Asks around — nobody's sure. Calls maintenance. 25 minutes pass. Tech comes, hits a reset sequence nobody on the line knew about. Same error Thursday. Same delay.
Downtime · cumulative00:00 → 00:30
30min · for a 2-min fix
With Mary 10:47 AM · Tue · routine
Knowledge that compounds.
"Getting error 14." Mary: "Sensor timeout. (A) Reset — hold START 3 sec then CLEAR, (B) Check proximity sensor on left rail." Operator taps A. Machine runs. Thursday, a different operator gets Error 14. Mary already knows.
Downtime · cumulative00:00 → 00:03
3min · added to knowledge base
These routine events happen 3–4× per week at a typical plant. Each one costs 20–45 minutes while the escalation chain runs. That's 8–12 hours of lost production a month — on problems that should've taken five minutes.
No separate documentation step. The interaction IS the record.
Mary works across your floor
Stamping PressesCNC Machining CentersInjection MoldingDie CastingForging PressesSpot WeldingAssembly LinesGrinding & FinishingHeat Treatment FurnacesWire EDMLaser CuttingHydraulic PressesSurface Mount (SMT)Paint & Coating LinesPackaging LinesStamping PressesCNC Machining CentersInjection MoldingDie CastingForging PressesSpot WeldingAssembly LinesGrinding & FinishingHeat Treatment FurnacesWire EDMLaser CuttingHydraulic PressesSurface Mount (SMT)Paint & Coating LinesPackaging Lines
FanucSiemensMitsubishiAllen-BradleySchulerMazakDMG MoriHaasKomatsuABBKUKAEngelArburgOmronBeckhoffBosch RexrothYaskawaHeidenhainTrumpfAmadaFanucSiemensMitsubishiAllen-BradleySchulerMazakDMG MoriHaasKomatsuABBKUKAEngelArburgOmronBeckhoffBosch RexrothYaskawaHeidenhainTrumpfAmada
Any machine with a control panel. No sensors. No hardware. No IT project.
How it works

Four steps. Five seconds to scan.

QR code on the machine. Phone in the operator's pocket. No new dashboards, no training program, no IT project.

01 / Trigger
Machine acts up.
Alarm, unusual noise, bad parts, an unexpected stop.
02 / Open
Operator opens Mary.
QR scan or dropdown. Describe the problem by voice or text.
03 / Guide
Mary guides.
Tappable A/B/C options. Right things tried in the right order.
04 / Resolve
Resolved or escalated.
Fixed → logged. Not fixed → maintenance gets the full context.
The numbers

Why this is a $50 billion problem.

The math behind the problem.

Problem scale
25
downtime events per plant, per month
Source · Siemens, 2024
Knowledge crisis
2.8M
US manufacturing workers retiring by 2033
Source · Deloitte
Cost
$50B
lost annually to unplanned downtime in US manufacturing
Source · Fortune Business Insights
Mary's ROI
7–18×
return on every dollar spent on Mary
Source · Based on pilot data
Who's behind this

Built by the team behind 200K+ industrial IoT deployments.

Co-founder · CEO
Aniket Deb
Built Bizongo — B2B supply chain platform, $300M+ raised. IIT Bombay. A decade of scaling complex B2B operations across manufacturing and supply chains.
LinkedIn ↗ aniket@flosync.io
Co-founder · CTO
Pushkar Limaye
Built Carnot Technologies — 200K+ IoT deployments, 7 patents, acquired by Mahindra & Mahindra. IIT Bombay. 11 years turning messy industrial data into systems that work.
LinkedIn ↗ pushkar@flosync.io
Currently piloting with leading auto-component manufacturers.
Status · Active pilots · 2026
Common questions

What people ask before the first call.

She escalates — but with context, not chaos. Maintenance gets a package: what the operator described, every step already attempted, photos from the session, and Mary's narrowed diagnosis. They start from a specific problem, not square one. And when they fix it, that resolution feeds back into Mary — so next time, the operator handles it alone.
No. Mary works with your existing machines and your operators' phones. QR code on the machine, web app on the phone. No sensors, no edge devices, no IT project. No app store download — works in any phone browser.
We start by loading 10–15 fault patterns per machine — curated from your manuals, your maintenance logs, and interviews with your best technicians. From day one of the pilot, every real event adds to the library. In three months, Mary knows patterns that aren't in any manual.
ChatGPT doesn't know that Station 12's feed assembly jammed three times last month. It can't escalate to your maintenance team with context. It doesn't learn from your specific events. Mary is grounded in your machines, your history, your plant. Generic AI gives generic answers.
One machine, three weeks. We build the fault library from your actual machines and your team's knowledge. You see real events flowing through Mary within the first week. At the end: a clear picture of resolution rates, time saved, and knowledge captured. Then you decide.
Book a call

See Mary in action.

30 minutes. We'll show you Mary on a real machine, walk through how it fits your plant, and answer questions. No pitch deck.

BringOne downtime story from your plant. We'll show you exactly how Mary would handle it.
Walk away withA clear picture of where Mary fits, ROI math for your plant, and pilot terms if it's a fit.
Prefer email? pushkar@flosync.io
Pick a time · 30-min slot · Pacific / Eastern
Loading available times…