AquaMind
The Intelligent Brain for Aquaculture and Water Treatment
AquaMind is WaterDoctor's AI-powered expert decision-making system. It integrates AquaOS data with deep expertise in water quality management, fish disease diagnosis, microbiology, engineering operations, and system maintenance — turning real-time telemetry into trustworthy, citation-grounded decisions.
What AquaMind does
AquaMind transforms experience-based aquaculture management into a data-driven, predictive, and standardized decision-making process.
Risk Prediction
Early-warning forecasts for ammonia accumulation, nitrite spikes, oxygen deficiency, disease outbreaks, feeding abnormalities, and biofilter instability.
Actionable SOP Generation
Context-aware standard operating procedures for treatment, feeding, and maintenance — tailored to each farm and pond.
Smarter Operational Decisions
Multi-specialist reasoning that weighs water quality, biology, nutrition, and engineering — with the trade-offs spelled out.
Real-time Monitoring & Analysis
Continuously ingests AquaOS telemetry, correlating signals across parameters to surface what matters.
Standardized Operations
A consistent, auditable, data-driven workflow that scales across sites, shifts, and species.
Cross-disciplinary Knowledge
Water quality, fish disease diagnosis, microbiology, engineering operations, and system maintenance — synthesized into every recommendation.
Early-warning detection
Examples of the risks AquaMind catches early — before they turn into losses.
AquaMind’s Position in the WaterDoctor Technology System
AquaMind is WaterDoctor’s third core IP. Built on top of the AquaOS data platform, it serves as the intelligent decision-making and expert-system layer. By combining real-time AquaOS data with domain knowledge in water quality management, fish disease diagnosis, microbiology, engineering operations, and system maintenance, AquaMind helps users move from seeing data to making better operational decisions.
- IP-1SND BacteriaFunctional microbial engine — system foundation
- IP-2AquaOSAI-IoT platform — data acquisition and operational base
- IP-3AquaMindIntelligent decision brain — expert system and risk prediction
- IP-4AI Digital Expert24/7 customer service and knowledge interface
- IP-5AquaChainTraceability, trusted records, and ESG data connection
How AquaMind works
A real team architecture — not a wrapper. Hierarchical down, peer across, verified through, delivered out.
Global Router (4-way)
Every case is triaged into clarify · agent-only · agent + expert review · or skip-to-human. The router never just answers — it routes.
T1 / T2 / T3 Supervisors
Three supervisor tiers coordinate specialists across aquaculture (T1), research (T2), and engineering (T3) — each with its own track of domain experts.
Domain specialists
Vision · Fish-Disease Dx · Microbiology · Water-Quality / RAS · Nutrition · Regulatory · Lit Review · Methodology · Process Designer · Equipment Selection · Cost/ROI — and more.
Shared workspace, not a pipeline
Specialists publish to a shared workspace; downstream specialists subscribe. The orchestrator convenes a Virtual Round Table per case.
Grounding & Citation Verifier
Final-pass authority. Every external claim is DOI-checked. Weakly-sourced claims are stripped or escalated to a PhD.
Delivery paths
Low-stakes → deliver. High-stakes → PhD signs. Skip-to-human → routed to the PhD pool.
System architecture
Six layers. Built for tenancy from day one. tenant_id is mandatory at every layer — from delivery channels down to the LLM substrate.
- 1Delivery Channels
Web portal · Partner REST · MCP server · White-label embed · AquaOS-embedded chat
- 2Agent Layer
Global router · Track supervisors · Specialists · Citation verifier
- 3Tool / Action Layer
Read tools · Write tools (policy-gated)
- 4Platform Services
Knowledge platform · Grounding · Marketplace · Eval harness · Audit log · Multi-tenancy · Identity / RBAC
- 5Substrate
LLM router (Claude / GPT / Gemini) · Vector DB · Temporal · Observability
- 6Integrations
AquaOS · LIMS · CrossRef / PubMed · FAO / regulators · Stripe Connect
AquaMind embedded in AquaOS
Real telemetry flows from ponds through MQTT into the Water Doctor backend, then directly into AquaMind. Customers don't switch apps — they ask AquaMind from inside AquaOS.
TRIGGER · GROUND · VERIFY · LEARN · EMBED
Sensor anomaly auto-loads last 7d telemetry
This pond's baseline — not a textbook range
Treatment Monitor watches DO / NH4 for 48-72h
Confirmed outcome feeds risk-window models
Bilingual ZH + EN chat lives in the AquaOS pond view
Example outputs
AquaMind ships three classes of deliverables — each fully cited, each tier-appropriate.
Fish-disease diagnosis with citation chain
Ranked candidate causes with confidence scores, a 5-horizon action plan, and a citation chain from peer-reviewed journals to internal handbooks — awaiting inner-ring PhD sign-off for high-stakes cases.
Journal-grade research with every DOI resolved
Methodology-reviewed structured reviews. Every reference is DOI-resolved — zero fabricated citations, zero vendor docs, zero news. Premium tier methodology reviewed by Dr Wang.
PFD + mass-balance + spec + ROI
Process flow diagram, mass-balance summary, equipment specification, cost/ROI, and standards compliance — delivered in days with the same auditability as a SGD 60k consultancy retrofit.
Three real cases
From morning panic to a cited treatment in 30 minutes — across the operator, the researcher, and the engineer.
From morning panic to a cited treatment in 30 minutes
Mr Chen, 6-pond tilapia farm, Guangdong. Vendor visit = SGD 800 and 2 days; wrong call = full pond loss. AquaMind delivers a Dr-Wang-signed treatment in 30 minutes — saving SGD 4-12k per incident.
Three weeks of literature work, done in a day
Dr Lin, PhD candidate, environmental engineering. Vanilla ChatGPT hallucinated 3 references. AquaMind delivers 47/47 DOI-resolved, 0 fabricated, methodology signed by a named PhD — overnight.
A SGD 60k consultancy retrofit — for SGD 8k in 4 days
Ms Goh, operations manager, pork-processing plant, Tuas. PUB tightened discharge 2026. Same audit trail, same PE-stamp, 14-month payback shown — 80-90% saving vs SGD 45-78k consultancy quotes.
Trust, governance & verification
AquaMind is not a generic chatbot. Every claim that goes out is grounded. Every external claim is DOI-checked. Weakly-sourced claims are stripped or escalated. High-stakes outputs are signed by an inner-ring PhD.
Every external claim cited and DOI-verified before delivery.
PhD sign-off gates all high-stakes outputs. Skip-to-human always available.
tenant_id mandatory at every layer; full audit log per case.
Promotion gate — 5% canary armed before new behavior rolls out.
Validated performance
AquaMind beats vanilla frontier LLMs (Claude, GPT, Gemini) on a blind-scored benchmark across three tracks — T1 Aquaculture Pro, T2 Env-Eng Research, T3 Industrial Water. Three SMEs per question across factual, citation, bilingual, refusal, and format dimensions.
Reproducibility paper co-authored by Dr Wang Chuansheng plus senior PhDs — public release on launch.
Integration with the WaterDoctor Ecosystem
AquaMind sits on top of the WaterDoctor stack — consuming data from every other IP and turning it into decisions.
