Cardano AI Mentor

[GENERAL] Name and surname of main applicant

Fitra Rahmani Khasyah (Full-stack Software Engineer) Khresna Pandu Izzaturrahman (Machine Learning & LLM Engineer)

[GENERAL] Are you delivering this project as an individual or as an entity (whether formally incorporated or not)

Individual

[GENERAL] Requested funds in ada

147435

[GENERAL] Please specify how many months you expect your project to last (from 2-12 months)

8

[GENERAL] Please indicate if your proposal has been auto-translated into English from another language

No

[GENERAL] Summarize your solution to the problem (200-character limit including spaces)

We will create a Cardano AI Mentor that uses semantic search and LLMs to provide accurate, real-time answers, centralizing all Cardano-related information on one user-friendly platform.

[GENERAL] Does your project have any dependencies on other organizations, technical or otherwise?

Yes

[GENERAL] If YES, please describe what the dependency is and why you believe it is essential for your project’s delivery. If NO, please write “No dependencies.”

Claude 3.5 Sonnet is an essential dependency for our project as it provides state of the art Large Language Model capabilities.

[GENERAL] Will your project’s output/s be fully open source?

Yes

[GENERAL] Please provide here more information on the open source status of your project outputs

Our project outputs will be released under the MIT License. This permissive license allows for broad usage, modification, and distribution of the software, encouraging community collaboration and innovation. This approach aligns with our commitment to open-source principles and ensures that the benefits of the Cardano AI Mentor project are accessible to all.

[METADATA] Horizons

AI

[SOLUTION] Please describe your proposed solution

To address the challenges of accessing scattered information in the Cardano ecosystem, we propose building a Cardano AI Mentor—a chatbot capable of providing accurate and up-to-date answers on any Cardano-related topic. This solution integrates advanced semantic search and LLM-based question-answering to deliver a seamless user experience on a single platform. Below, we detail the technical components and implementation plan:

1. Data Collection and Storage

• Data Crawling: We will set up pipelines to crawl and extract information from multiple sources, including Cardano forums, official websites, and community platforms like Reddit. This process ensures we capture a comprehensive view of the latest developments, user discussions, and official announcements.
• Sources include:
• Cardano Forum
• Cardano.org
• Reddit Cardano Community
• Cardano Ideascale
• Vector Database: Extracted data will be stored as vectors in Milvus, a high-performance, open-source vector database. Milvus allows us to perform similarity searches, which are crucial for finding relevant information based on user queries. The use of Milvus ensures that our system can handle large-scale data and maintain low-latency query responses.

2. Backend and Chat Implementation

• Backend Development: The backend will be built using Node.js to manage interactions between the chatbot and the semantic search service. It will handle authentication, session management, and API calls to the LLM. Additionally, we will conduct a database schema design to optimize data storage and retrieval processes, with PostgreSQL used to store chat histories that can be leveraged for continuous model refinement.
• Frontend Development: The user interface for the chatbot will be developed using React, providing a responsive and intuitive experience. This interface will enable users to interact directly with the AI mentor, submit queries, and receive contextualized answers.

3. Model Training and Implementation

• Semantic Search Model: We will train a semantic search model that understands user queries more than simple keyword matches. By using state-of-the-art models such as sentence-transformers/all-mpnet-base-v2 and google-bert/bert-base-uncased, our system will accurately interpret user intent and retrieve the most relevant content from the vector database. As new models become available, we will continue to evaluate and adopt those that offer best performance.
• LLM Integration: The LLM component, based on Claude 3.5 Sonnet, will handle generating user-friendly, conversational answers by using the relevant content retrieved through semantic search. This ensures that the user receives accurate, detailed, and easy-to-understand responses.
• Retrieval-Augmented Generation (RAG): We will implement a RAG pipeline to combine the strengths of semantic search and LLM-based generation. This process involves:
• Accepting user input and converting it into a vector representation.
• Using the semantic search model to find the most relevant context from Milvus.
• Feeding the retrieved context into the LLM to generate a response that is both informative and accurate.

4. Deployment and Hosting

• Server Setup: We will configure our servers to meet the needs of training the model (using Google Colab Pro) and running the live service using VPS. This setup enables effective resource allocation while ensuring optimal performance for user demands.
• Public Rate-Limited API: A publicly accessible, rate-limited API will be implemented for inference, allowing developers to integrate the Cardano AI Mentor into their applications while managing usage.

5. Continuous Improvement and Training

• Chat Data Storage: The PostgreSQL database will store user interactions with the chatbot, creating a valuable dataset for further fine-tuning and training of the model. By analyzing these interactions, we can identify gaps in the model’s understanding and iteratively improve its accuracy and relevance.
• Training and Model Updates: As new data becomes available or as the Cardano ecosystem evolves, we will periodically retrain the semantic search model to ensure it remains up-to-date and relevant. This ongoing training process helps to maintain a high-quality user experience.
• Testing: We will conduct comprehensive testing towards the end of the project timeline to ensure system functionality, accuracy, and user satisfaction.

[IMPACT] Please define the positive impact your project will have on the wider Cardano community

The Challenge: Fragmented Information on Cardano

Currently, information about Cardano is dispersed across various platforms, making it difficult to stay updated. For instance, to track ongoing projects in the Catalyst ecosystem, users must visit cardano.ideascale.com. To access official releases and documentation, cardano.org is the primary source. Additionally, community-driven discussions and rumors often require searching through Reddit. This fragmentation makes it challenging to find the latest updates on Cardano, even when the needed information exists—it is not always easy to locate or access specific details.

While tools like ChatGPT or Claude can assist with general inquiries, their inherent knowledge cut-off means they cannot provide the most current data or updates. As a result, relying solely on these models further limits access to up-to-date information.

Impact on User Engagement

This scattered nature of information creates significant barriers for both new and existing users. Newcomers interested in learning about Cardano may become discouraged if they struggle to find the information they seek, potentially halting their exploration of the ecosystem. Similarly, existing users may find it challenging to stay engaged if keeping up with updates becomes too cumbersome. This situation can ultimately hinder user engagement and community growth, as users may become less active or choose to disengage altogether.

Our Proposed Solution: Cardano AI Mentor

To address the challenges of fragmented information outlined above, we propose the development of Cardano AI Mentor, a chatbot designed to provide users with instant access to the latest information on any Cardano-related topic. This AI-powered mentor will enable users to ask any question they have about Cardano, and our system will search for the most relevant information needed to answer their queries. The answers will then be generated using a state-of-the-art Large Language Model (LLM), making it possible to access all the necessary information about Cardano from a single platform.

Key Features and How They Drive Impact

1. Semantic Search for Accurate Responses: A key component of the Cardano AI Mentor is the semantic search model, which allows the system to understand the meaning behind user queries rather than just matching keywords. This enables more nuanced searches, where even if a user’s wording doesn’t match the stored content exactly, the system can still retrieve relevant results. For example, if a user asks, "What is the roadmap of Cardano?" the semantic search model will compare it with related content like "The roadmap of Cardano is..."
2. Up-to-Date Information Through Real-Time Data Collection: The AI Mentor keeps its knowledge base current by using real-time data crawling scripts that extract and aggregate information from Cardano’s core platforms, including Cardano Forum, cardano.org, Reddit, and IdeaScale. This ensures users always receive the latest developments, avoiding the outdated knowledge limitations found in models like ChatGPT.
3. Centralized User Interface for Simplified Access: All Cardano-related information will be consolidated into a single, user-friendly interface, dramatically reducing the time and effort required for users to find relevant information. Rather than visiting multiple sites or forums, users can access everything they need in one place, creating a seamless search experience that encourages deeper engagement.
4. Public Chat API for Broader Ecosystem Impact: In addition to the web-based interface, the project will provide a public Chat API, enabling developers and other projects within the Cardano ecosystem to integrate the chatbot’s capabilities into their own applications. This API will extend the value of the mentor beyond the immediate platform, multiplying its utility across the ecosystem.

Impact on the Cardano Ecosystem

The Cardano AI Mentor is not just a solution to fragmented information; it has the potential to increase user engagement within the Cardano ecosystem significantly. By making it easier for both new and existing users to access up-to-date information, the mentor can help onboard new users smoothly.

• Newcomers who might have been discouraged by the effort required to find information will have easy access to comprehensive resources, transforming curious observers into active participants.
• Existing users will also benefit by staying effortlessly informed about the latest developments, projects, and community discussions, encouraging continued engagement and reducing user churn.

This increased engagement will contribute to a more vibrant, connected Cardano community, creating a stronger ecosystem. The centralized interface, combined with advanced semantic search and real-time data integration, ensures that information access is smooth, reducing participation barriers and attracting new users to become long-term contributors of the platform.

In summary, the Cardano AI Mentor will be a game-changer by:

• Leveraging semantic search models to deliver highly relevant answers.
• Providing real-time information using data crawlers to keep the knowledge base fresh.
• Centralizing all relevant Cardano information into one easy-to-use platform.
• Offering a public Chat API to extend the AI mentor’s reach and impact across the ecosystem.

These features will ensure that both new and existing users can engage with the Cardano ecosystem more easily, leading to sustained growth and participation across the community.

[CAPABILITY & FEASIBILITY] What is your capability to deliver your project with high levels of trust and accountability? How do you intend to validate if your approach is feasible?

Our profile:

Khresna Pandu I

Linkedin profile: https://www.linkedin.com/in/khresna-pandu/

Github profile: https://github.com/KhresnaPanduI

Educational Background & Technical Expertise:

• Computer Science Degree: Graduated from Gadjah Mada University, the top university in Indonesia, with a specialization in deep learning.
• Thesis Project on Computer Vision: Developed a computer vision project focused on Indonesian Sign Language Recognition using MediaPipe. The project is available on GitHub: Link to Thesis Project.
• In-Depth Knowledge of AI Models: Proficient in deploying and fine-tuning state-of-the-art large language models (LLMs) such as OpenAI’s GPT series, Claude, LLaMA, and Mistral. This expertise allows Khresna to lead the integration of LLMs and semantic search for this project.
• Proven Track Record: Proven track record: As a Core LLM Engineer in his previous company, Pandu successfully led projects that are available to 3+ millions of users, directly impacted business operations, working closely with C-level executives and department heads. This demonstrates Pandu’s ability to handle high-stakes projects with accountability.
• Technical expertise: Experience in fine-tuning and deploying various LLMs, both open source (LLaMA, Mistral, Mixtral7x8) and 3rd party hosted models (OpenAI and Claude). Optimized these models for performance, increasing inference speed by 300% and reducing memory usage by 73%.
• Data management proficiency: Pandu worked with large datasets, improving corpus data quality by 70% for a 12,000-record database, with less than 30ms latency for each query. Pandu also set up and managed vector and relational databases capable of handling 2+ TB of data.
• Data crawling: Pandu have developed multiple data scraping and crawling pipelines, processing over 10 gigabytes of data daily. This expertise is important for gathering large volumes of Cardano-related data from online sources.
• Scalability testing: Given his experience in handling high-volume requests (500+ per minute for sentiment analysis API, 50+ simultaneous requests for LLM inference), Pandu will design and conduct scalability tests to ensure the system can meet Cardano's needs.
• Real-world application testing: Similar to how Pandu developed a chatbot that reduced response times by 99.5%, Pandu will set up real-world scenarios to test the system's effectiveness in the context of Cardano's specific use cases.
• Active involvement in the Cardano community: Contributing as proposal reviewer since Fund 8.

Fitra Rahmani Khasyah

Linkedin profile: https://www.linkedin.com/in/fitra-rahmani/

Github profile: https://github.com/khasyah-fr

Proven track record: Fitra graduated from Gadjah Mada University with a computer science degree and has a paper published about Reinforcement Learning (a category of machine learning) (https://link.springer.com/chapter/10.1007/978-3-031-29927-8_8). As a software engineer, Fitra has developed services and projects using various tech stack such as Node.js, Golang and PHP, orchestrating over 150,000 daily transactions data. He designed and managed large PostgreSQL databases, Redis clusters, and NGINX setups in his previous work within the e-commerce industry. He’s been involved within Cardano’s Project Catalyst since Fund 8 as a Reviewer and has been granted a funding before from the SingularityNet (Decentralized AI Blockchain) platform (https://deepfunding.ai/proposal/singularitynet-and-deep-funding-indonesia-chapter/) showcasing his project management skills and committed interests for blockchain technologies. Fitra will handle the development of the front-end service which handles the user interactions as well as the back-end service that would orchestrate the authentication, session management, websocket connections, and AI model inference.

We believe our approach is extremely feasible because we have researched the tech stack and implemented the fundamental concepts needed for this project:

Crawling source: 

• https://forum.cardano.org/
• https://cardano.org/
• https://www.reddit.com/r/cardano/
• https://cardano.ideascale.com/c/campaigns/

Backend Tech Stack:

• Node.js
• Milvus
• Postgresql
• Docker

Front End Tech Stack:

• React
• Web Socket

Semantic Search Model:

Semantic search allows us to find relevant information by understanding the meaning behind user queries rather than just matching exact words. Unlike traditional relational databases that rely on precise keyword matching, semantic search focuses on the underlying meaning, similar to how Google retrieves relevant web pages.

For this project, we will use state-of-the-art models to achieve accurate semantic understanding. Currently, popular choices include models like sentence-transformers/all-mpnet-base-v2 and google-bert/bert-base-uncased. However, as the field evolves rapidly, we will continue researching and adopting the best available models.

Vector Database: Milvus

To perform semantic search, data must be stored as vectors, which allow us to measure similarity between user queries and stored information. This is where vector databases come in. Among the various options, we have chosen Milvus, an open-source, high-performance vector database. It is widely used by organizations like NVIDIA, PayPal, Roblox, and LINE, making it a reliable choice for our needs.

LLM Model:

We aim to use the best available model in the market. As this field evolves rapidly, the top choice at the time of this proposal is Claude 3.5 Sonnet. We will utilize its API, which costs $3.50 per million input tokens and $15 per million output tokens. However, we remain flexible and will consider adopting newer models if more advanced options become available.

[PROJECT MILESTONES] What are the key milestones you need to achieve in order to complete your project successfully?

Total Timeline: 8 months (approx. 32 weeks)

Project Setup and Data Collection

Duration: 5 weeks

Tasks:

• Project Management Setup: Set up project management tools (Notion) for tracking tasks so the community can monitor the project’s progress.
• Repository Setup: Create a GitHub repository with version control and access management. Set up the project’s dedicated GitHub account, control access, and repository.
• PostgreSQL Database Schema Design & Setup: Design the PostgreSQL database schema for efficient data storage and retrieval.
• UI Design: Create initial flow for the user interface, focusing on how the user interacts with the site. Curate low-fidelity design materials.
• Data Crawling & Storage: Develop and test scripts to extract data from Cardano Forum, Cardano.org, Reddit, and Ideascale. Then storing it in PostgreSQL.

A: Milestone Outputs:

• PostgreSQL database with schema and data storage.
• GitHub repository initialized.
• Successfully crawled data from at least 3 Cardano-related sites.
• Notion tracking board in use.

B: Acceptance Criteria:

• Collecting data from 3 different data sources.
• UI design validated for development.
• All tasks tracked in Notion.

C: Evidence of Milestone Completion:

• PostgreSQL is populated along with its schema.
• GitHub repository accessible.
• Screenshots of the UI designs.
• Notion available for the community to view.
• API test results demonstrating successful ingestion.

Initial System Development & Semantic Model Training

Duration: 6 weeks

Tasks:

• Initial Front-End Development: Begin initial front-end development using React, setting up the UI components and layout for the Cardano AI Mentor site.
• Initial Back-End Development: Start developing back-end services to support initial functionalities, such as request routing and data retrieval.
• Semantic Search Model Training: Perform multiple training iterations to refine the semantic search model. Using google-bert/bert-base-uncased  and sentence-transformers/all-mpnet-base-v2 model.
• Model Evaluation: Adjust hyperparameters and test with sample queries to improve accuracy.

A: Milestone Outputs:

• Initial front-end service developed, consisting of UI layouts and components.
• Initial back-end service implemented, supporting early functionalities like routing and data retrieval.
• Trained semantic search model deployed and ready for internal testing.
• Model evaluation report with results from sample queries.

B: Acceptance Criteria:

• Completed initial layouts and components for the front-end service.
• Successful data retrieval and routing from the back-end service.
• Initial semantic search model achieves at least 70% query accuracy in internal testing.

C: Evidence of Milestone Completion:

• Front-end code repository containing the layouts and components.
• Back-end code repository containing the routing and data retrieval.
• Code repository with trained models and hyperparameter configurations documented.
• Model evaluation report comparing performance metrics from different iterations.

Initial Core System Implementation and Data Infrastructure Setup

Duration: 6 weeks

Tasks:

• Milvus Setup: Configure columns, schema, and optimize Milvus for efficient semantic search.
• Data Conversion Script: Convert PostgreSQL data into vectors for Milvus ingestion.
• User Authentication & Session Management: Implement user authentication and secure session management for chat access.
• WebSocket Integration: Implement the WebSocket module for real-time communication between the front end and back end.

A: Milestone Outputs:

• Milvus instance operational with optimized schema and ready for semantic search.
• Data conversion script deployed and successfully converting PostgreSQL data into vectors.
• User authentication and session management module implemented for controlled chat access.
• WebSocket module implemented and working for real-time communication.

B: Acceptance Criteria:

• Milvus configured to handle multiple concurrent searches.
• Authentication and session management correctly handle user access, with secure session termination and renewal.
• WebSocket supports real-time communication.

C: Evidence of Milestone Completion:

• Code repository containing the data conversion script and API documentation.
• Ingestion reports confirming successful conversion and insertion of PostgreSQL data into Milvus.
• Back-end code repository containing the implementation of authentication and session management, as well as WebSocket real-time communication.
• Milvus configuration logs showing operational status and optimized schema setup.

Data Pipeline Development and Final Core System Implementation

Duration: 6 weeks

Tasks:

• Preprocessing Scripts: Develop scripts to split multi-question queries into sub-questions for better search results.
• Prompt Engineering: Explore prompt strategies (e.g., few-shot learning, ReAct) to optimize responses from the LLM.
• Additional Data Crawling: Add more Cardano-related sites to increase data coverage.
• Chat API Development:Implement the chat pipeline: user input → semantic search → Claude → user-friendly response.
• API for Semantic Search: Develop and integrate the API for semantic search with the backend.
• Front-End & Back-End Integration: Work on integrating the front-end and back-end systems to enable seamless interaction flow.

A: Milestone Outputs:

• New Cardano-related sites integrated into the data crawling process.
• Integrated front-end and back-end interaction flow established.
• A functional chat API pipeline allowing seamless interactions from input to response.

B: Acceptance Criteria:

• Preprocessing scripts split 90% of multi-question queries accurately.
• At least 2 additional Cardano-related sources crawled and ingested into the database.
• Chat API pipeline passes end-to-end testing, returning responses within 2 seconds for the first word.
• Front-end and back-end integration achieves smooth interaction flow with no broken components.

C: Evidence of Milestone Completion:

• Back-end code repository containing the semantic search response handling.
• Demo or screenshots showcasing integrated front-end and back-end functionality.
• API test results demonstrating that the chat pipeline functions correctly.
• Code repository containing deployed preprocessing scripts with test cases.
• Logs or reports showing successful crawling from new data sources.

Deployment and MVP Release

Duration: 6 weeks

Tasks:

• Final Back-End Development: Complete the back-end, focusing on implementing the load testing and rate-limiting feature using NGINX to ensure system stability under high traffic.
• System Monitoring and Maintenance: Implement Prometheus + Grafana to monitor the system in real-time and set up alerts for issues.
• Deployment Configuration: Configure the server for deployment, ensuring hosting, infrastructure, and environment setups are ready for production usage.
• MVP Release: Launch the initial version, gather feedback from the community, and perform usability testing.
• Bug Fixes and Refinements: Address issues from community feedback.
• Model Refinement and Edge Case Handling: Re-train and fine-tune the semantic search model based on feedback and new data.

A: Milestone Outputs:

• MVP available for public user testing.
• Back-end infrastructure ready with load testing completed and NGINX rate-limiting implemented.
• Prometheus + Grafana system operational with real-time monitoring and alerts enabled.
• Re-trained semantic search model deployed with improvements.

B: Acceptance Criteria:

• Usability testing completed with actionable feedback documented.
• Deployment completed with no critical infrastructure issues.
• Model refinement based on community input achieves at least a 5% increase in query accuracy for edge cases.

C: Evidence of Milestone Completion:

• NGINX configuration logs confirming successful rate-limiting and load management.
• Monitoring dashboards and alert configurations available in Prometheus and Grafana.
• Usability testing report with feedback from at least 10 community users.
• System logs showing successful handling of multiple concurrent user sessions without major downtime.

Final Launch and Maintenance

Duration: 3 weeks

Tasks:

• Final Launch: Deploy a fully optimized version with a stable backend and UI.
• Documentation and Knowledge Transfer: Create technical documentation for handover.
• Project Closeout Video and Community Updates: Publish a summary video and engage the Cardano community.
• Community Updates: Provide updates to the Cardano community through forum posts, social media, and other channels, sharing project progress and results.

A: Milestone Outputs:

• Fully deployed and optimized version of the Cardano AI Mentor.
• Technical documentation completed and available for handover to stakeholders and the community.
• Project closeout video published and shared with the Cardano community.

B: Acceptance Criteria:

• The AI Mentor platform is live and accessible to the public with no major downtime reported.
• Community updates cover all milestones and provide insights on the project’s success and future roadmap.

C: Evidence of Milestone Completion:

• Project Closeout Report and Video (PCR and PCV) shared across Cardano community channels (Telegram & Discord)
• Screenshots and post links showcasing updates shared across social media and forums.

[RESOURCES] Who is in the project team and what are their roles?

Fitra Rahmani Khasyah - Software Engineer

Fitra will focus on these aspects of the project:

• Front-end development: Implementing the front-end website UI for users to interact with using React and integrate with the backend.
• Back-end development: Developing the authentication, session management, and calls to the trained LLM model.
• Database and system design: Provisioning the database schema and indexing for scalable backend queries as well as integrating NGINX and the server.

Khresna Pandu Izzaturrahman - ML & LLM Engineer

Pandu will focus on key technical aspects of the project, including:

• Data Crawling: Designing and implementing data scraping pipelines to collect Cardano-related data from various sources.
• Model Fine-Tuning and Training: Fine-tuning the semantic search model to optimize its performance using the collected data and managing the training process.
• AI Integration: Integrating the LLM for generating accurate, context-driven responses.
• API Development: Creating API services for semantic search and chat functionalities, ensuring smooth interactions between users and the system.

[BUDGET & COSTS] Please provide a cost breakdown of the proposed work and resources

Using ADA Rate of 0.35

This means 1 USD equals 2.857 ADA.

Domain Name

For the website, we need to book a domain name and we aim to provide the front-end service for 3 years.

• Cost: $20/year for 3 years = $60
• In ADA: 60 * 2.857 = 171 ADA 

Google Clab Pro

To train the semantic search model, we need access to a high-specification GPU. A Google Colab Pro subscription will be required throughout the project for continuous model testing and refinement.

• Cost: $10/month for 8 months = $80
• In ADA: 80 * 2.857 = 228 ADA
• Reference: Google Colab Pro Pricing

Server for the front-end and back-end application

For the application, we expect the workload to be medium. To keep the costs optimal, we opt for a VPS instead of cloud solutions such as AWS, Google Cloud, or Azure.

• Platform: Contabo Cloud VPS (Tier 3)
• Cost: $17.5/month for 36 months = $630
• In ADA: 630 * 2.857 = 1800 ADA
• This ensures the AI Cardano Mentor front-end and back-end can run for 3 years.

Server for Semantic Search

The semantic search engine requires a dedicated server to handle user queries, as it performs all inference locally. We will use the same VPS specification for this purpose.

• Platform: Contabo Cloud VPS (Tier 3)
• Cost: $17.5/month for 36 months = $630
• In ADA: 630 * 2.857 = 1800 ADA

This VPS will also be maintained for 3 years.

Claude API Costs

The API costs for Claude are competitive, with rates of $3.50 per million input tokens and $15 per million output tokens. Given that retrieved search results are counted as input, each message can be relatively large in terms of token usage. We are assuming each message have 1,000 tokens input and 200 tokens output.

Cost per input token = $3.50/1,000,000 = $0.0000035 

Total cost for input token = 1,000 tokens x $0.0000035/token = $0.0035 

Cost per output token = $15/1,000,000 = $0.000015 

Total cost for output token = 200 tokens x $0.000015/token = $0.003 

Total price per message= $0.0065 

Assuming we got 100 messages/day for 3 year straight:

Total Claude cost = $0.0065 x 100 x 365 x 3=711.75 $ 

Total Claude cost in ADA = 711.75 x 2.857 ADA = 2033 ADA

Personnel Cost

Both Pandu and Fitra will dedicate 20 hours per week to this project, with a competitive rate of $35/hour for 8 months.

• Calculation:
• 20 hours/week × $35/hour × 2 persons × 32 weeks × 2.857 ADA = 128,000 ADA

The $35/hour rate reflects the average rate for experienced professionals on Upwork, ensuring fair compensation for skilled developers. Both Pandu and Fitra have a proven professional track record within their field and have been involved within Cardano’s Project Catalyst for a long time, adding reliability to this project.

References:

• Full-Stack Developers on Upwork
• AI Engineers on Upwork

Total Fixed Cost

Domain Name: 171 ADA

Google Colab Pro: 228 ADA

Claude API: 2033 ADA

Web Server (VPS for Hosting): 1,800 ADA

Model Server (VPS for Semantic Search): 1,800 ADA

Personnel Cost: 128,000 ADA

Total Fixed Cost:

171 + 228 + 1,800 + 2033 + 1,800 + 128,000 = 134,032 ADA

Contingency (10% of Fixed Cost)

Based on our experience, ADA price fluctuation is one of the most prominent issues when it comes to successfully delivering a proposal. Therefore, we mitigate this by adding a contingency fund, which reduces the price fluctuation and unexpected costs risk.

Contingency = 10% × 134,032  ADA = 13,4032 ADA

Final Cost

Final Cost = 134,032 ADA + 13,4032  ADA = 147,435 ADA

[VALUE FOR MONEY] How does the cost of the project represent value for money for the Cardano ecosystem?

The Cardano AI Mentor project represents exceptional value for money for the Cardano ecosystem for several key reasons:

1. Advanced Semantic Search Capabilities: The project leverages a state-of-the-art semantic search model to provide highly relevant answers to user queries. This technology ensures that users receive accurate, context-aware responses, significantly improving the quality of information retrieval compared to traditional keyword-based systems.
2. Real-Time, Up-to-Date Information: With the implementation of efficient crawling scripts, the AI Mentor constantly updates its knowledge base with the latest information from various Cardano sources. This ensures that users always have access to the most current data, news, and developments within the ecosystem.
3. Centralized User Interface: The project creates a single, user-friendly interface that consolidates information from across the Cardano ecosystem. This centralization dramatically reduces the time and effort required for users to find relevant information, improving overall ecosystem engagement and knowledge dissemination.
4. Public Chat API: By offering a public Chat API, the project extends its value beyond the user interface. Developers and other projects within the Cardano ecosystem can integrate this powerful tool into their own applications, multiplying the impact and utility of the initial investment.
5. Cost-Efficient Development: The personnel costs, at $35/hour, represent exceptional value given the expertise required. This rate is significantly lower than typical market rates for senior full-stack developers and AI engineers, which can range from $60-$100/hour or more. This efficient pricing ensures high-quality output while maintaining reasonable project costs. References: Full-Stack Developers on Upwork, AI Engineers on Upwork
6. Economical Infrastructure Choices: The use of VPS for hosting, at just $17.5/month, is highly cost-effective. Comparable AWS instances could cost up to $270/month, representing potential savings of over 93%. Utilizing Google Colab Pro for model training at $10/month is extremely economical. An equivalent AWS GPU instance could cost up to $365/month, showcasing savings of over 97%. References:AWS Server AWS Server with GPU
7. Long-term Information Accessibility: For an investment of 147,435 ADA, the Cardano community gains a powerful, AI-driven platform that centralizes and simplifies access to ecosystem information. This tool will serve the community for years, significantly reducing the time and effort required for both new and existing users to find accurate, up-to-date information about Cardano.
8. Community Growth Catalyst: By lowering the barrier to entry for new users and enhancing engagement for existing community members, the AI Mentor has the potential to accelerate ecosystem growth. The easier access to information can lead to increased participation in Cardano projects, more informed voting in Catalyst, and overall stronger community involvement. This growth potential far outweighs the initial investment.
9. Scalability and Future-Proofing: The modular design and use of adaptable technologies like semantic search models mean that the system can be easily updated and expanded as the Cardano ecosystem grows. This scalability ensures that the initial investment continues to provide value even as the ecosystem evolves.
10. Open-Source Nature: By making the project outputs open-source under the MIT License, the investment benefits not just the immediate users of the AI Mentor, but also allows for community contributions and potential spin-off projects, multiplying the value generated from the initial funding.

In conclusion, the Cardano AI Mentor project offers substantial value for money by providing a long-term, scalable solution to information accessibility challenges in the Cardano ecosystem. Its combination of advanced features like semantic search and real-time data updates, coupled with highly cost-efficient development and infrastructure choices, positions it as a high-impact investment for the future of Cardano. The centralized interface and public API further extend its utility across the ecosystem. When viewed against the potential for ecosystem-wide benefits, long-term usage, and the significant cost savings in development and infrastructure, this project represents an excellent use of community funds that will deliver value far beyond its initial cost.