Continuous audit intelligence for maritime

Maritime audit is a leading indicator.
Treat it like one.

safyrmind is a maritime-vertical AI platform that turns every audit, inspection, and observation into a live signal — anticipating risk before it becomes a detention, a claim, or a lost charter.

0
Regulatory
regimes covered
0
Risk dimensions
continuously scored
24/7
Agentic engine
never sleeps
SafyrScore™ · Fleet risk
42-vessel tanker fleet
Moderate
0.0
▼ 3.2 vs 90d ago
P
68
D
82
K
71
B
76
C
63
Trigger fired · T-411
Pre-arrival PSC readiness
Aegis Titan → Rotterdam · Paris MoU · elevated target factor · 72h out.
90-DAY TREND ↓ IMPROVING
The problem

The maritime industry measures audit outcomes.
It should be anticipating them.

Every fleet operator carries a growing operational-risk envelope — vetting failures, PSC detentions, P&I claims, insurance surprises. Traditional QHSE tools capture findings after they happen. Institutional knowledge walks off with every retiring officer. The audit rhythm has not evolved for the scale, complexity, and stakes of today's shipping.

01 · The recording problem

Findings are captured. Learnings are not.

Ten similar findings across a fleet do not become a systemic signal. They become ten CAPAs, closed independently, with the same gap reopening on the next vessel.

0%
of tanker findings recur within 24 months on comparable hulls
02 · The prediction problem

Detentions and vetting failures arrive as surprises.

The signals that predict failure exist across telemetry, records, and behaviour — but they sit in silos. No one sees the vessel deteriorating until the inspector does.

$50k–$2M
single-event cost of a detention or a lost charter, per vessel
03 · The financial-risk problem

The CFO cannot see what the audit programme sees.

Operational-risk exposure is one of the largest classes of loss a shipping company carries — and it is invisible to the finance function until an event materialises. Insurance renewals, charter approvals, and reinsurance strategy all depend on data no one is producing.

Zero
continuous financial view of audit-driven operational risk in most fleets
How it works

A continuous learning loop,
not a static checklist.

safyrmind doesn't run on a static checklist someone updates by hand. It runs on Safyrmind MARIQ™ — a purpose-built, continuously growing knowledge base of audit questions drawn from governing regulations, standards, and industry requirements — paired with a domain-specialised model that reasons about every finding the way an experienced auditor would, and a proprietary scoring methodology that turns compliance data into one defensible risk number. The system gets sharper with every validated finding, not just every release.

1
Safyrmind MARIQ

The question corpus

Automatically generates a comprehensive, context-aware corpus of audit questions from regulatory and industry requirements — not a static checklist someone updates by hand.

Corpus growing, continuously
2
Safyrmind-LLM / CAPA Forge

Finding interpretation

Interprets every finding using domain-specific maritime knowledge and relevant historical records, then predicts the finding-specific corrective and preventive action.

Rust streaking, lifeboat davit, Fr. 42
→ Re-coat and inspect davit pins fleet-wide
3
Safyrmind SafyrScore

Quantified vessel risk

A proprietary risk-scoring methodology quantifies vessel risk from findings, CAPAs, and relevant vessel and operational factors — one defensible number.

Score 71
One number, fleet-wide
Our mission

To make maritime audit intelligence continuous, defensible, and decisively actionable.

Every audit becomes a live signal. Every finding becomes a fleet-wide learning opportunity. Every risk becomes anticipated, not recorded. We build the platform that turns audit from a compliance overhead into the shipping industry's earliest and clearest indicator of operational risk.

Our vision

A shipping industry where risk is anticipated — and expertise is at everyone's hand.

A world where every seafarer, from Master to rating, has the intelligence of an experienced peer beside them. Where no detention arrives unforeseen, no charter is lost to a preventable finding, and no institutional knowledge is lost when an officer retires. That is the industry safyrmind is built to serve.

A DPA reviewing fleet risk on a wall-mounted dashboard in a shore office overlooking the port
From vision to fleet reality

The same SafyrScore™ that guides an officer on the bridge rolls up into the view the DPA sees ashore — one number, understood the same way at every level.

Who we serve

Built for every role that carries
the weight of the vessel.

safyrmind is designed for the concentric rings of accountability in a shipping company — onboard command, shore management, and the finance function that translates operational risk into commercial reality.

Onboard command

Master, C/O, C/E

One question: what needs your authority today? Audits arrive in flow, on the tablet, with a copilot who thinks like a peer.

−53%Time on audit
admin per week
3Priority cards
on Monday
VoiceWake phrase
copilot
Master's command view · Monday 06:12 UTC
3 ITEMS
Sign
IG O2 analyser calibration · closure ready
F-2402
Trigger
Pre-arrival Yosu · PSC readiness
T-411
Upcoming
SIRE 2.0 · Yosu berth 2 · 62% pass prob
T-2d
Weekly hours on audit admin · 8 weeks
From 18 h/wk to 8.5 h/wk · same audit rigour, less paperwork
Shore accountability

DPA & QHSE Manager

SMS integrity as a defensibility question. Every signal groundable to a record; every finding on the same category becomes a programme-level review.

−58%Finding recurrence
fleet-wide
1→5Systemic signal
escalation
AutoMonthly DPA
report draft
Finding recurrence rate · fleet-wide, 12 months
▼ 58%
54% 18%
Systemic signal fired at M4 · fleet-wide L4 intervention deployed · recurrence dropped through M12.
Internal audit programme · 12-month completion
Marine
92%
Technical
88%
Safety / ISM
96%
Navigation
84%
Annual VDR
78%
Fleet operations

Fleet Director & TS

Which vessel needs you most this week? Ranked by real-time risk, with the calendar 12 weeks ahead and the visit briefing pre-drafted.

−34%Visit time
on wrong vessels
12wkInspection
horizon
Top 10Auto-drafted
visit briefing
Hot-risk top 5 · fleet of 42
3 IN ELEVATED
1
Aegis Titan
78
2
Aegis Meridian
74
3
Aegis Sceptre
69
4
Aegis Voyager
65
5
Aegis Herald
61
Inspection calendar · next 12 weeks
W1W6W12
5 inspections in horizon · Titan PSC W4 flagged for pre-arrival readiness workflow.
Commercial readiness

Vetting Coordinator

No vessel walks into a SIRE 2.0 headed for failure. Continuous chapter-level readiness with mock inspections and predicted pass probability.

+18ppSIRE pass
probability
14 dPre-inspection
prep window
MockLLM as
inspector
SIRE 2.0 readiness · Aegis Meridian · Yosu T-14d
62% PASS PROB
78%
Ready
Ch.1
89
Ch.6
64
Ch.9
82
Ch.10
76
Ch.11
78
Pass probability trend · 90 days to inspection
65% threshold
Crossed 65% floor at T-42d after Ch.6 gap remediation · projected 82% at inspection.
Finance function

CFO & Risk Manager

Operational risk quantified. Expected Annual Loss with confidence bands, live detention cost, insurance-renewal prep in one view.

P50/ P75 / P95
bands on every $
LiveDetention cost
hourly ticker
6.1×Aggregate ROI
on L4 spend
Fleet Expected Annual Loss · 12 months
▼ 34%
$3.8M $2.5M
P50
P75
P95 band
Loss class composition · Q1
Vetting 31%
Det 22%
P&I 18%
H&M
Env
Detentions down 3→0 · P&I claim frequency down 12% · vetting failures down 86%.
Two navigation officers on the bridge of a polar-class vessel, radar and ECDIS displays lit, icebergs visible through the windows
Every rank, one platform

From the officer of the watch to the DPA ashore — the same intelligence, scoped to what their rank needs to see.

The platform

Seven layers.
One continuous intelligence.

safyrmind is built as a maritime vertical LLM sitting on an agentic engine, a system-of-record intelligence layer, and a cryptographically-anchored evidence fabric. Every layer is deployable independently — together they form a continuous intelligence loop from sensor to sign-off.

Layer 1 · The agent

Event-driven agentic decisioning

The right audit fires to the right rank at the right moment — driven by age, event, prediction, campaign, and systemic signals.

5Trigger classes
always running
<2 mTriage to
dispatch latency
AutoSystemic-signal
escalation
Agent activity · 24 hours · fleet-wide
14 FIRED · 12 CLOSED
3:12·T-401 age
6:47·T-408 OT
11:20 · T-411 pred
0006121824
Age
Event
Predictive
Campaign
Scheduled
Systemic signal detected at 11:20 · same finding on 5 vessels in 60 d → programme-level review escalated to DPA queue.
Layer 2 · The sensors

Real-time hardware intelligence

OT telemetry, VDR, CCTV, and IoT flow into the risk model in real time — with vessel-side segregation preserved per IMO MSC.428(98).

50kEvents / sec
steady state
Anomaly detection
threshold
IMOMSC.428(98)
segregation
Main-engine cyl 3 exhaust temp · rolling 60 min
ANOMALY · SCOPED AUDIT FIRED
+2σ −2σ +3.8σ · T-408
Spike at T−4 min triggered scoped audit to C/E on watch · closed loop through L1 → dispatch → tablet in ECR.
Layer 3 · The surfaces

Multi-surface front-end

Native tablet, mobile, web, mixed-reality — each engineered for its operating context. Offline-first, wet-hand tolerant, one-hand safe.

4Surfaces
native each
30 dOffline
runtime
1-handMobile thumb
zone safe
Audit volume by surface · last 30 days
2,847 SESSIONS
📱
Tablet
42%
📲
Mobile
31%
💻
Web
21%
🥽
MR
6%
Onboard delivery mode split
Online · VSAT 58%
Low-bw 27%
Offline 15%
Every capture works offline · queued sync drains in priority order when link returns.
Layer 4 · The learning loop

Continuous training & learning

Every finding becomes a 3-minute rank-specific micro-module. Every intervention measured 90 days out — dollars saved per dollar spent.

24 hFinding → module
fleet-wide
6.1×Aggregate
ROI · L4
12Working
languages
IG O2 finding rate · L4 module deployed M0
▼ 74%
Month 0 · pre-intervention
4.2
findings / vessel-quarter
Month 3 · post-intervention
1.1
findings / vessel-quarter
Top 5 modules by measured ROI
IG O2 cal
9.0×
Enclosed space
8.3×
MLC rest hrs
6.7×
PMS Ch.6
5.3×
ECDIS familiarity
2.4×
Layer 5 · The dashboard

360° dashboard with cause-effect

A working surface, not a compliance report. The SafyrScore™ composite decomposes to five dimensions and nine regime lenses. Every panel resolves to an action.

5Risk
dimensions
9Regime
lenses
PinAny panel
per persona
Fleet dashboard · shore user landing
LIVE
Fleet score
42.3
Moderate · ▼ 3.2 · 90 d
Hot risk · top N
5
Titan · Meridian · Sceptre
Findings open
63
39 within SLA · 24 breached
Triggers 24 h
14
12 auto-closed · 2 to review
Dimension breakdown · fleet composite
P
68
D
82
K
71
B
76
C
63
Layer 6 · The copilot

In-audit knowledge referencing

A maritime peer at hand — clause interpretation, consequence differentiation by port MoU, sourced to the original text. Voice-first onboard.

<2 sResponse
latency (p95)
100%Answers cited
to source
"safyr"Voice wake
phrase
Copilot · vetting coordinator asks Ch.6 · Q14
GROUNDED · 3 SOURCES
What does SIRE 2.0 Ch.6 Q14 actually mean, and what evidence would satisfy it?
Intent: the inspector is checking that the IG O2 analyser is capable of producing a compliant reading — calibrated within the interval required by SOLAS Ch.II-2 Reg.4.5.10 and the manufacturer.

What good looks like: cal record ≤ 90 d, live reading matching cal standard within tolerance, in-date cal-gas cert, and consistent OT stream over last 60 min.
📘 SOLAS Ch.II-2 · 4.5.10 📘 SIRE 2.0 · Ch.6 · Q14 🗂 Fleet history · 12 mo
What happens if we fail this at Yosu?
Regime-specific consequences · Tokyo MoU + active fire-safety CIC · detention risk quantified · prior history cited.
Layer 7 · The vision

Multi-modal inference

Reads gauges, documents, and photos. Transcription is the error; safyrmind removes it. Every capture cross-checked and confidence-banded.

4Modalities
read natively
≥ 90%Confidence bar
for sufficiency
CrossCal-gas + OT
stream verify
L7 inference · IG O2 analyser · Manifold deck
SUFFICIENT
O₂ %
20.7
Inferred reading
20.7% ± 0.03
Confidence94%
✓ Cal-gas cert on file
✓ OT stream steady 20.6-20.8
✓ Reading within tolerance
Layer 8 · The output

Automated reporting cadence

Weekly digests, monthly reviews, quarterly board packs, closure reports — auto-drafted, cryptographically anchored, ready to defend.

4Formats
every report
RFC3161
timestamps
SHAManifest per
evidence pack
Report cadence · fleet · rolling 12 months
100% ON TIME
Weekly digest
52 / yr
Monthly review
12 / yr
Quarterly board
4 / yr
Closure reports
on demand
Evidence packs
per event
Every artifact PDF · DOCX · XLSX · ZIP · with SHA manifest + RFC 3161 timestamp.
Bulk carrier stern showing bow thruster and tug-push markings
Fleet-wide, not fleet-by-fleet

Nine regulatory regimes, one evidence trail — mapped automatically to every vessel in the fleet.

Coverage

Nine regulatory regimes.
One clause-level knowledge base.

Every audit question in Safyrmind MARIQ™, every evidence requirement, and every finding is tagged to one or more of nine regimes. The same underlying evidence propagates across every regime where it is admissible — captured once, satisfies many.

SIRE 2.0

OCIMF tanker vetting

13 chapters, three question tiers. Pre-inspection preparation, mock inspections, chapter-level readiness scoring.

RISQ 3.2

RightShip

Fleet-wide risk screening across tanker, bulk, and specialised tonnage. Continuous readiness, never point-in-time.

PSC + CIC

Port State Control

Paris, Tokyo, IO, Caribbean, Med, Black Sea, Abuja, Viña del Mar, Riyadh MoUs + USCG. CIC campaign preparation built-in.

Class

Class societies

DNV, LR, ABS, BV, ClassNK, KR, CCS, RINA, IRS. Annual, intermediate, special surveys — plus enhanced survey programme.

Flag

Flag state

Marshall Islands, Panama, Liberia, Singapore, Malta, Bahamas, HK, UK, Cyprus, Greece — each flag's emphasis modelled.

P&I

Loss-prevention surveys

IG club condition surveys, entry surveys, thematic loss-prevention. Findings routed through CAPA Forge™.

ISO Integrated

Environmental & H&S

ISO 9001, 14001, 45001, 50001 as one integrated management system. Single audit event, multi-standard evidence.

ISM

Internal audits

Rolling 12-month programme per IMO Res. A.741(18). Auditor-independence verified. DPA sign-off queue.

Internal · 5

Company-specific

Marine, Technical, Safety, Navigation, Annual VDR — five programmes, own rhythm, own ownership, own question banks.

Cross-regime map · A rest-hour log satisfies MLC in the Flag audit, crew management in RISQ 3.2, competence in ISM Element 6, and Ch. 10 crew questions in SIRE 2.0 — all from one capture. safyrmind maintains the map so evidence propagates automatically.
Ship captain using a tablet in the wheelhouse at sea
Continuous, not point-in-time

The audit runs whether or not the inspector is on board — and whether or not the link is.

Before & after

Twelve months on the platform.
Here is what the same fleet looks like.

A mid-size tanker operator — 28 vessels, average age 11 years, mixed trade — modelled at rollout and again 12 months later. The score, the trend, and the metrics that drive them. No editorial gloss; the numbers that changed.

Month 0 · Baseline

SafyrScore™ — fleet risk

Rising trend · reactive audit programme
68.5
Elevated
▲ 4.8 vs prior year
12-MONTH TREND DETERIORATING
Detentions · 12 mo
3
≈ $540k direct cost
Vetting failures
7
Charter losses on 2 hulls
Finding recurrence
54%
Same category, new vessel
Median CAPA close
62 d
Against 21-day SLA
Month 12 · Post-rollout

SafyrScore™ — fleet risk

Improving trend · continuous anticipation
34.2
Moderate
▼ 34.3 vs baseline
12-MONTH TREND IMPROVING
Detentions · 12 mo
0
2 predicted & averted
Vetting failures
1
86% reduction
Finding recurrence
18%
Systemic signals closed
Median CAPA close
19 d
Within SLA · 69% faster
Illustrative composite based on typical outcomes across comparable tanker operators in the first 12 months. Model the numbers for your own fleet in the ROI calculator below.
Return on investment

Model your operational-risk exposure.
See what safyrmind returns.

Every input below drives the same SafyrScore™ methodology in real time. Impact ranges are calibrated against IG Club and industry-benchmark data. The output is what a comparable fleet has actually saved in the first 24 months on the platform.

Your fleet profile

Adjust the sliders. The result updates as you move.

25 vessels
11 years
3 per vessel
Modelled outcome · 24 months

Your expected return

Annual operational-risk exposure avoided (P50)
$1.2M
P75 $1.9M · P95 $3.2M
Detentions avoided
1.2 / yr
≈ $180k cost avoided
Vetting pass rate lift
+18%
≈ $840k charter revenue protected
CAPA closure time
-42%
Median days-to-close
Aggregate ROI
6.1×
Loss reduction per platform dollar
Model calibrated on comparable fleet outcomes at the 24-month mark. Impact ranges anchored to IG Club claims data and MoU detention cost benchmarks. Individual results depend on execution — book a demo to see the model tuned to your specific vessel mix and trade lanes.
What operators are saying

Three roles.
Three vantage points on the same platform.

The onboard officer running an audit in flow. The DPA defending the SMS to an inspector. The CFO defending the operational-risk exposure to a board. Different questions, same continuous intelligence.

"

Cargo readiness used to be an evening of paperwork after a full day on deck. Now it happens as I walk — the tablet asks the right question at the right tank, the copilot handles the clause references, and evidence is captured as I capture it. The Master signs off on the same deck. The audit is the operation now.

Cargo audit time
-53%
Vetting observations
-71%
CO
Chief Officer
Aframax tanker · Northline Tankers
"

The systemic-signal detector changed how we run the SMS. When the same finding appeared on five vessels in sixty days, it stopped being five CAPAs and became one programme-level review. My monthly DPA report writes itself — grounded in specific records, every claim citable. I defend the SMS with evidence now, not with narrative.

Finding recurrence
-58%
ISM audit prep
-6 wk
DP
Designated Person Ashore
42-vessel product tanker fleet · Meridian Shipping Group
"

Operational risk used to be the largest number my board could not see. safyrmind turned it into a continuous view with confidence bands — the same discipline we bring to fuel exposure or freight risk. At the last insurance renewal, we walked in with three years of measurable improvement. Our premium held flat in a hardening market.

Detentions avoided
2 / yr
Renewal outcome
Flat
CF
Chief Financial Officer
Ship management group · Halcyon Marine Holdings
Illustrative composites drawn from typical outcomes across comparable operators. Names and organisations are hypothetical; the numbers reflect real fleet trajectories.
Ship officer reviewing digital compliance documents on a tablet on the bridge
ON THE BRIDGE — WHERE THE AUDIT ACTUALLY HAPPENS
About safyrmind

Built by seafarers.
Engineered by AI experts.

Founded by Master Mariners and software engineers who bring together two kinds of expertise rarely found in the same team: deep maritime operating experience and deep technical expertise in AI, LLMs, and enterprise software.

The wider team combines Master Mariners, ex-vetting inspectors, DPAs, technical superintendents, applied AI researchers, and enterprise engineers. That combination gives safyrmind both the domain depth to understand the problem and the engineering depth to solve it at the system level.

Our engineering founders have built software and technology at some of the world's leading technology companies, with expertise spanning AI and LLM architectures, model adaptation, inference, and intelligent software systems. Our maritime founders have commanded ships, conducted audits, defended findings, and lived the operational consequences the platform is designed to anticipate.

We are building a maritime-first AI platform — not a general-purpose LLM with a shipping wrapper. Maritime knowledge, finding patterns, regulatory context, operational workflows, and rank-specific decisioning are embedded throughout the system.

We don't just apply AI to maritime.
We combine maritime expertise with the engineering depth to build AI that understands how the industry actually works.

Domain-native

Built from real audits, real findings, real operational decisions, and decades of maritime experience. The domain is the source of truth, not simply a target market.

AI-native

Deep engineering expertise across LLM architectures, RAG, LoRA/PEFT, model adaptation, inference, evaluation, and production AI systems.

Evidence-grounded

Recommendations are grounded in clauses, records, evidence, and modelled probabilities with explicit confidence — not unsupported AI opinions.

Command-respecting

AI assists; accountable officers decide. Sign-off authority remains with the person responsible for the operation.

Trust & security

Built for regulated
enterprise scale.

Encryption

AES-256 at rest, TLS 1.3 in transit, BYOK KMS.

Data residency

EU · US · APAC regions — you choose primary and DR.

Full audit trail

Actor, timestamp, before / after — every mutation.

Model lineage

Every output traces to inputs, corpus, and version.

An officer reviewing fleet data on a tablet at a container terminal gangway, a dockworker visible behind
Evidence you can stand behind

Every score, every flagged finding, traces back to the audit that produced it — on deck, in port, wherever the work happens.

Ready to see
your fleet's audit intelligence?

Book a 30-minute demo. We will show the platform running on a comparable fleet to yours — with the risk model, the copilot, and the CFO view tuned to your operational reality.

Talk to the team