LiveLive on first asset class — 126,958 chunks · 13,652 images · 171 case studies

Every Engineer On Shift
Is Your Best

AssetBlue captures how the best engineers reason — and turns that judgment into a living system that can diagnose, investigate, guide, and teach across every asset and every shift.

app.assetblue.ai / investigation
Ranked Hypotheses
Boiler Tube Leakage · Unit 4
Economizer · BHEL 210MW · 1993 · Last OH: 8mo
#1Fly ash erosion — economizer, gas-facing
87% High confidence
#2Overheating from partial blockage
52% Probable
#3Soot blower impingement
28% Possible
The problem

The gap isn't information. It's engineering judgment.

Plants already have alarms, manuals, maintenance logs, and SOPs. What they don't have is a scalable way to preserve and apply the reasoning that turns signals into decisions.

$0T
Lost annually to unplanned downtime across the world's 500 biggest companies
0%
Of organisations consistently capture knowledge from departing retirees
0K+
Americans reaching age 65 each day from 2024 through 2027
$0M/hr
Cost of downtime in a large automotive plant
0–47%
Of maintenance organisations still operate without CMMS
0–7 days
Average time to complete a root cause analysis. Manually.
From the field

If the same problem happened at another thermal plant, we don't have access to that knowledge. We call NTPC, call BHEL, wait for callbacks.

Executive Engineer
Boiler Maintenance

A comprehensive list of failures of that particular equipment worldwide — that would be helpful.

Efficiency Engineer
Energy Manager

We have a history book only. If he has not faced such problems, he can go through that — he'll get some idea.

Senior Engineer
On how new engineers learn
Why this hasn't been solved

Industrial reasoning breaks traditional AI.

Complexity explodes

Industrial systems are highly interconnected. As variable count rises, the space of possible causal structures grows too quickly for naive discovery.

Failure data is sparse

Critical assets don't fail often enough to generate rich training data. Real-world failure events are rare, uneven, and incomplete.

Black boxes don't earn trust

Classification models can label likely causes, but they can't show the reasoning path, causal mechanism, or evidence chain behind the answer.

Expertise doesn't scale

Traditional RCA methods like 5 Whys and FTA are powerful, but they remain manual, slow, and dependent on scarce experts.

Sources: APQC · Siemens 2024 · ABB 2023 · UpKeep 2024 · RTPS Field Research 2025
Beyond post-failure

Not just when things break. When things don't look right.

The same reasoning engine that supports post-failure diagnosis helps engineers investigate anomalies before they become shutdowns.

Examine & Investigate
"This bearing sounds different."

An engineer opens AssetBlue, describes what they're seeing or hearing, and shares a photo or video. The system reasons through possible causes, retrieves similar patterns from prior cases, and helps the engineer investigate before the issue escalates.

The physical world does not speed up. Thinking can.
Anomaly InvestigationInvestigating
ObservationUnusual high-frequency vibration on ID Fan B
Visual evidence📷 2 photos uploaded · bearing housing
Similar patterns3 prior cases retrieved · 2 from same asset class
System suggestionCheck lubrication schedule · Compare vibration baseline
How it works

Structured reasoning. Not keyword search.

AssetBlue guides engineers through structured investigation using Fishbone, 5 Whys, Fault Tree, and FMEA — not as static templates, but as live reasoning paths.

Phase 01: Intake & Context
New Investigation
Equipment
Boiler Unit 4 · Economizer
Symptom
Steam leak detected during inspection
🎤 Voice
📷 Photo
⌨️ Text
Coach mode

Diagnose and learn. Simultaneously.

Most systems automate workflow. AssetBlue develops judgment. Every interaction transfers reasoning — not just answers — so the workforce gets stronger with every investigation.

What to observe
What to rule out
Why each question matters
Symptom → mechanism → fix
“It's not working.”
“Fly ash erosion, gas-facing tubes, high-ash coal exposure.”
How it compounds

Every session makes it smarter.

Wrong diagnoses are auto-tagged, classified, and reviewed weekly. The knowledge base and the workforce improve together — automatically.

Session 1Possible
3 questions · 8 min
Session 10High confidence
1 question · 2m 41s
Session 20Opens with hypothesis
0 questions · < 2 min

The switching cost isn't data migration. It's the relationship.

Validation

Live on the first asset class. Measured against reality.

Structured evaluation against 171 expert-validated failure cases across 9 industries. Each case scored on five dimensions against a plain-LLM baseline.

+0%
Knowledge base lift over plain-LLM baseline
0%
Cases where knowledge base wins or ties
0
Root cause accuracy (KB-augmented)
0
Failure mechanism match (KB-augmented)
Corpus scale
126,958
Document chunks indexed
13,652
Diagnostic images classified
171
Structured case studies
146
Failure modes mapped
220
Operating envelopes
57
Asset component nodes
Evaluated across
InstitutionalUtilitySteelFood ProcessingPulp & PaperChemical ProcessingMetal ProcessingOil RecoveryCustom Engineering
Where the KB matters most

On the hardest cases — unusual mechanisms, cross-industry failures, sparse prior documentation — the knowledge base turns misses into diagnoses.

Pulp & paper case example
Baseline: 0.10
KB: 0.64
Ground-validated with Karnataka Power Corporation (KPCL).
Built for you

Different role. Same conviction.

Your diagnostic partner at 2am.

AssetBlue works the way you work — voice-first, shift-aware, and built for the field.

Start in under 60 seconds

Open the app. Select the asset. Describe what you see. Investigation begins.

🔍

Works before failures, not just after

Investigate anomalies, unusual sounds, and drift — not just breakdowns.

📚

Your plant's history, instantly searchable

Every past investigation feeds a queryable knowledge base specific to your site.

🔧

Built for the field

Voice-first. Glove-friendly. Works across shifts. No dashboard to navigate.

2m 41s
Average time to root cause by session 10
Platform

Six commitments. Non-negotiable.

Investigations persist across shifts

A Case survives shift changes, dead batteries, and interruptions. Any engineer can continue from where it stopped.

Case-based

Every output has a next step

Every hypothesis includes the action, the owner, the priority, and whether it's safe without supervisor sign-off.

Action-complete

Every claim cites its source

Traceable to a case study, OEM standard, or technical reference. Tappable. Disputable.

Cited and verifiable
🌐

Build once. Deploy across 32 industries.

41 asset types span $36T of industrial output. One boiler model works in power, mining, chemicals, and refining.

Universal assets

Adapts to the engineer automatically

Detects expertise and urgency in real time. Senior gets speed. Junior gets reasoning. No mode to select.

Automatic coaching

Designed for 2am, not 2pm

Voice input. Glove-friendly targets. No dashboard. The app opens to one screen: Start RCA.

Worst-shift tested
How we compare

Purpose-built beats general-purpose.

STATUS QUO
Excel · WhatsApp · Tribal Knowledge
  • 3–7 days per investigation
  • Knowledge locked in heads
  • No cross-plant learning
  • No audit trail
  • Repeat failures uninvestigated
70–75% of facilities
ENTERPRISE APM
IBM Maximo · GE · SAP
  • ₹2–10 crore implementation
  • 12–18 month deployment
  • Requires data science team
  • RCA is a checkbox feature
  • Predicts failures, doesn't diagnose them
15–20% of enterprises
ASSETBLUE
Diagnostic Intelligence
  • Root cause in hours, not days
  • Knowledge captured and queryable
  • Cross-plant learning from day one
  • Cited evidence on every output
  • Complements Maximo/SAP — not a replacement
Purpose-built. 1/10th the cost. 1/10th the time.
EASY TO COPY
  • ·Conversational UI wrapper around an LLM
  • ·Basic RAG pipeline over technical PDFs
  • ·Generic troubleshooting chatbot
  • ·Mobile form-factor for field teams
HARD TO REPLICATE
  • Curated, semantically indexed failure corpus per asset class
  • Structured causal reasoning over domain-specific failure graphs
  • Coaching-first interaction calibrated to engineer expertise
  • Compounding knowledge from every completed investigation
  • Field-validated across real plant conditions and languages

Others monitor. Others document. Others route service knowledge. AssetBlue is building the reasoning layer.

Our standard
>0%
RCA Correctness Rate. Per asset class. Per plant.

Not a target — a floor. A wrong diagnosis acted on once means the engineer never opens the app again. We validate against physical outcomes, not satisfaction surveys.

Correctness over speed

A fast wrong answer destroys trust permanently.

Cited, never asserted

Every output traces to a case study, standard, or validated pattern.

Human-reviewed improvement

Nothing deploys without expert sign-off.

60-day recurrence tracking

If the same failure recurs, the diagnosis was wrong — and the system learns.

Founding research team

Six researchers across computer vision, NLP, edge AI, and applied ML — spanning IIIT Hyderabad and industry.

Advisory Board
C. V. Jawahar
Professor & Amazon Chair, Dean R&D, IIIT Hyderabad. INAE Fellow. Founded CVIT.
28,146 citations
Shailesh Kumar
Chief Data Scientist. PhD UT-Austin. 20+ patents. ex-Google Brain.
5,355 citations
Sandeep Kumar Shukla
Director, IIIT Hyderabad. IEEE Fellow, ACM Distinguished Scientist.
8,785 citations
Core Researchers
Manish Shrivastava
LTRC, IIIT Hyderabad. Multilingual NLP.
4,320 citations
Vineet Gandhi
CVIT, IIIT Hyderabad. PhD INRIA. Visual saliency.
1,570 citations
Priyesh Shukla
CVEST, IIIT Hyderabad. PhD UIC. ex-Samsung. Edge inference.
Asset coverage

One engine. Every critical asset.

🔥
Industrial Boiler
LIVE
🌀
ID Fan
LIVE
💨
FD Fan
LIVE
⚙️
Steam Turbine
2026
🔌
Transformer
2026
🏗
Compressor
2026
Generator
2026
🔥
Industrial Boiler
LIVE
🌀
ID Fan
LIVE
💨
FD Fan
LIVE
⚙️
Steam Turbine
2026
🔌
Transformer
2026
🏗
Compressor
2026
Generator
2026
🔄
Heat Exchanger
2026
💧
Centrifugal Pump
2026
🔧
Electric Motor
2026
🛢
Pressure Vessel
2026
Crusher
2026
🔩
Gearbox
2026
📡
Control Valve
2026
🔄
Heat Exchanger
2026
💧
Centrifugal Pump
2026
🔧
Electric Motor
2026
🛢
Pressure Vessel
2026
Crusher
2026
🔩
Gearbox
2026
📡
Control Valve
2026

Questions from every plant visit.

Selecting pilot partners — power, oil & gas, manufacturing, mining & minerals

Describe the problem.
Let's reason through it together.

Request a demo. We'll walk you through a live diagnostic session in 30 minutes. Or request a pilot implementation with your asset class.

Or email us at info@assetblue.ai

From the team that built Embibe — India's largest AI education platform, backed by Reliance Jio.