Aquanautical
#DeeperOceanResearch
The problem Aquanautical solves
The Problem Aquanautical Solves
1. Inaccessible Deep-Sea Knowledge
- The deep sea is Earth’s largest ecosystem, yet it remains the least studied.
- Collecting and analyzing biodiversity data requires expensive expeditions, lab equipment, and expert taxonomists.
- Schools and many researchers have no access to live, authentic data for learning or discovery.
2. Fragmented & Incomplete Databases
- Existing species databases are incomplete and often outdated.
- Novel organisms detected by eDNA or cameras remain unclassified because tools rely on reference data that doesn’t exist.
- This creates a blind spot in biodiversity monitoring and conservation.
3. Barriers in Education & Engagement
- Biology and AI in classrooms are taught with static textbooks or toy datasets.
- Students rarely get hands-on exposure to real scientific workflows, let alone contribute to discovery.
- This leads to disengagement and missed opportunities to train the next generation of AI-literate scientists.
What People Can Use Aquanautical For
For Researchers & Conservationists
- Automated Taxonomy: Classify eDNA sequences and images into species (or higher taxa) faster and with confidence scoring.
- Abundance Mapping: Visualize population trends and species distribution in real time.
- Novelty Discovery: Detect potential new species via unsupervised clustering and anomaly detection.
- Conservation Alerts: Receive warnings about pollution spikes, invasive species, or biodiversity loss.
- Scalable Storage & Access: Centralized, queryable biodiversity data with versioning and provenance.
For Education (Schools & Universities)
- Interactive Learning: Students explore live ocean data instead of memorizing textbook charts.
- Hands-on AI Experience: Classes use real DNA and image datasets to understand how AI works in biology.
- Gamification: Badges, leaderboards, and challenges make learning science engaging.
- Collaborative Discovery: Students can cluster unknown DNA sequences or label new organisms, contributing to global science.
- Curriculum Integration: Lesson packs align with biology, ecology, data science, and AI learning objectives.
For Policy Makers & NGOs
- Decision Support: Evidence-based recommendations for quotas, restoration, and protected areas.
- Transparent Conservation Funding: Blockchain-based tracking of donations and impact.
- Public Engagement: Share visualizations and discoveries with communities to build awareness.
How It Makes Tasks Easier / Safer
- Faster Research: Cuts weeks of manual taxonomy into automated minutes.
- Safer Fieldwork: Reduces the need for hazardous deep-sea expeditions by using eDNA and AI models.
- Accessible Education: Makes advanced biodiversity research tools usable by teachers and students anywhere, even without lab equipment.
- Reliable Monitoring: Provides real-time dashboards with 99.7% uptime — enabling proactive conservation actions instead of delayed responses.
- Democratized Discovery: Anyone from a student in a classroom to a marine researcher can explore, classify, and even discover new species through one unified platform.
Challenges I ran into
Challenges I Ran Into
1. Handling Incomplete eDNA Databases
The hurdle:
When running eDNA classification, many sequences didn’t match existing reference databases. This meant the model either misclassified them or returned “unknown”.
How I solved it:
- Integrated unsupervised clustering and novelty detection so unknown sequences could still be grouped meaningfully.
- Flagged these as “putative new taxa” instead of discarding them, turning a problem into a discovery opportunity.
2. Training Vision Models on Limited Data
The hurdle:
Deep-sea species images are rare and imbalanced — some species have thousands of images, while others have only a few. This caused the AI model to overfit to common classes.
How I solved it:
- Used data augmentation (rotation, low-light simulation, blur) to mimic deep-sea conditions.
- Applied transfer learning from general marine datasets and fine-tuned on smaller sets.
- Balanced training with class-weighted loss functions.
3. Real-Time Processing & Visualization
The hurdle:
Displaying live data streams (eDNA results, ocean parameters, species maps) in the dashboard without lag was tricky, especially when handling large batches of data.
How I solved it:
- Implemented WebSockets for real-time updates instead of slow polling.
- Optimized database queries with MongoDB indexing for fast retrieval.
- Batched heavy tasks into a queue system (Celery/Redis) so the UI always stayed responsive.
4. Making AI Outputs Understandable for Students
The hurdle:
Raw model probabilities and taxonomic data were too technical for educational use. Teachers and students needed simple, visual explanations.
How I solved it:
- Built confidence bars and taxonomy cards with plain-language explanations.
- Added voice assistant support so students could ask natural questions like:
“What family does this fish belong to?”
5. Ensuring System Scalability
The hurdle:
With multiple schools, researchers, and NGOs accessing the platform, scaling data pipelines without downtime was a big concern.
How I solved it:
- Adopted a microservices architecture with containerization (Docker/Kubernetes).
- Can used cloud storage for raw data and MongoDB for structured results.
- Built an admin retraining panel so models improve continuously without service interruptions.
Tracks Applied (1)
AI in Education
Technologies used
Discussion
Builders also viewed
See more projects on Devfolio