Analytics Working Group - Season 3 Proposal

Analytics Working Group - S3 Proposal

Proposal for Inverse Finance DAO to cover operations of the Analytics Working Group (AWG) in Season 3, running for 6 months from November 1, 2024 to April 30, 2025.

1. Introduction

This proposal outlines the agenda for Season 3 of the Analytics Working Group (AWG) within the Inverse Finance DAO. Leveraging the success and lessons of Seasons 1 & 2, we aim to deepen our data analytics capabilities to guide strategic decision-making and stimulate growth. For transparency purposes this proposal will be provided with detailed annexes about the AWG achievements last season (Annex I) as well as the cost involved (Annex II) and available on the (this) forum.

2. Analytics Working Group Overview

2.1 Role and Importance within Inverse Finance DAO

The AWG’s primary role is to analyze, interpret, and leverage data for strategic decision-making, through the provision of analytics & AI tools or infrastructures, thereby offering invaluable insights, identifying risks, growth opportunities, and highlighting improvement areas.

2.2 Achievements of Season 2

In Season 2, the AWG enhanced its operational excellence and technical resources by integrating Financial Machine Learning and an advanced Alert Monitoring system to its lead product : Inverse Watch.

Our day-to-day operations—developing data queries, producing reports, and managing data inquiries—remained strong. No major service disruptions have been observed in season 2.

Significant achievements include :

  • Maintaining and monitoring the stability and efficiency of our alerting systems 24/7 (inverse.watch and associated services) ;

  • Implementing a state of the art Financial Machine Learning training and predicting framework on Inverse Watch ;

  • Implementing a pilot functionality for Smart Contract indexing ;

  • Improving our advanced graph visualization mode on Inverse Watch ;

  • Implementing a dark mode on inverse Watch allowing our team to work longer hours more in a more pleasant manner ;

  • Improving the stability and error handling of our discord bots to continuously monitor prices or smart contract functions ;

  • Improving our blockchain simulation framework on Inverse Watch ;

  • Adding various dashboards relying on those systems including :

    • sINV monitoring ;
    • LP DOLA weight monitoring ;
    • Executive Report Indicators ;
    • Aave vs DBR rates historical comparison
  • Adding various alerts relying on those systems including :

    • Price Feeds monitoring
    • FiRM alerts on Twitter : foster engagement on social networks engagement

These initiatives improved our analytics, risk modeling, and made transactional data more intuitive, reinforcing the platform’s role in strategic decision-making for the DAO.

More details are available in the Annex I.

2.3. S2 Success Metrics

AWG is happy to report that all success metrics/objectives have been implemented successfully in season 2 :

  • :white_check_mark:Infrastructure Uptime
  • :white_check_mark:Frequency of insights reports publications
  • :white_check_mark:Project: Advanced Data Connectors and Visualization Tools
  • :white_check_mark:Project: Machine Learning and AI Capabilities
  • :white_check_mark:Project : Bots stability improvement
  • :white_check_mark:Project : Alert monitoring page More details are available in the Annex I.

3. WG Goals for Season 3

3.1 Strategic Objectives

For Season 3, the AWG is set to maintain its analytics capabilities, further refine its tools, and strengthen its collaboration with other working groups, aiming to create a more cohesive and data-driven DAO.

3.2 Usual Objectives

The main objective of the AWG is to continue operating and maintaining its systems “as-usual” in order to ensure their availability and stability and guarantee their effectiveness.

3.3 Additional Projects for Season 3

  • Treasury Revenue : coordinate with Zerion (or alt) to correct data and build reports to analyze historically Treasury operations and revenue.
  • Token Terminal integration : coordinate with Token Terminal to display inverse finance metrics on their website and boost growth & awareness.
  • Machine Learning : Leverage our newborn ML infrastructure to assist in the DAO decision making and Risk monitoring
  • Smart contract indexation : improve the embryonic indexing functionality implemented in season 2.
  • Dashboards : Expand the coverage of our dashboards to operational functions (ie assist Fed policy and Treasury decision-making)

3.4. Success Metrics for S3

Success in Season 3 will be quantified through metrics such as infrastructure uptime, the frequency of insight report publication, and the completion of the above listed projects.

3.5. Responsibilities and Focus Areas

  • Data Strategy Development and Governance : Our focus remains on refining our data strategy to ensure it aligns perfectly with the DAO’s goals. This involves comprehensive data management and governance, ensuring data integrity, security, accessibility and affordability.

  • Tooling, Infrastructure Maintenance and Improvement : The AWG is dedicated 24/7 to maintaining and improving the robustness of its platform and tools, ensuring they operate efficiently and are continuously improved to meet the DAO’s evolving needs.

  • Data Analysis, Interpretation, and Insight Provision : We will continue to utilize and improve our tools to delve deeper into data, offering clear, actionable insights that can significantly influence the DAO’s strategic directions.

  • Dashboard & Report Generation : The generation of detailed reports on various topics remains a priority. These reports play a crucial role in guiding risk decision processes in the DAO.

  • Alerting System and Automated Socials Enhancement : We aim to further refine our sophisticated alerting system, ensuring we remain vigilant and responsive to any operational or financial risks that may arise.

  • Collaboration, Consultation, and Data Literacy Advocacy: By enhancing our collaboration with other working groups, we seek to provide essential data support and promote a culture of data literacy across the DAO, empowering all members with the tools & knowledge to make data-informed decisions.

  • Other AWG Activities : Like other team members @naoufel is on standby 24/7 on the following Multi Sigs :

    • Fed Chair
    • Treasury Working Group
    • Policy Committee

4. Decision-making Power and Governance

The AWG will maintain its decision-making authority over analytics infrastructure and tooling. This autonomy ensures we can efficiently allocate resources in alignment with our strategic goals, driving forward our analytics capabilities.

5. SecOps Responsibilities and Enhancements

Our ongoing collaboration with the Risk and Treasury Working Groups highlights our vital role in providing the technical expertise and analytical support necessary for the DAO’s operational and risk management strategies.

6. Budget for Season 3

  1. S2 Budget summary and remainder

The budget for S2 originally included $4,748.4 with an additional $3,800, totaling $8,548. However, the total cost for season 2 amounted to only $4,041.27 which is notably lower than anticipated. This surplus could allow for coverage of expenses for several months beyond what was initially planned for season 2.

Detailed budget considerations for tooling, infrastructure maintenance, and potential project expansions will be provided separately, ensuring transparency and accountability in the allocation of DAO resources.

  1. Contributors
Name FTE Band Pro-Rata Monthly Salary Total for Season 3
Naoufel 1 B 12,000 77,000
  1. Ad-hoc & Tooling
  • dRPC: ~100 $ / month
  • Digital Ocean : ~500$/month
  • GRT Tokens: $1,200 for S2 (unused in S1 & S2)
  • OpenAI : ~ 0 to 50 $/month
  • Coingecko: ~156 $ / month
  1. Flexible Budget

AWG requests a flexible allowance of 1,000 DOLAs to cover tooling unforeseen expenditure that arises during the Season.

  1. Summary
S3 $DOLA allowance S3 $$INV allowance
Contributors 77,000 0
Ad Hoc & Tooling 6,036 0
Flexible Budget 1,000 0
Total 84,036 0

10. Conclusion

The AWG’s achievements in Season 2 have laid a robust foundation for scaling our analytics communications, analysis, and reporting in coverage and complexity with a sophisticated approach.

With Season 3, we aim to build upon this success, building innovative reports and analysis to enhance decision-making, risk management and operational efficiency across the DAO.

Annex I : S2 Achievements

Infrastructure Uptime

  • Uptime Performance: The AWG maintained an infrastructure uptime of over 99.5%, as verified by our previous monitoring provider, StatusPage. This high level of reliability underscores our commitment to ensuring platform availability for the DAO’s critical analytics and monitoring services.
  • Monitoring Improvements: To further enhance monitoring capabilities, the AWG has initiated the transition to Better Stack, a more advanced tool for status and log monitoring. Better Stack offers improved data granularity and uptime tracking for our virtual machines (VMs), ensuring more accurate and accessible uptime and downtime metrics. This upgrade is expected to bolster our ability to detect and respond swiftly to any system issues, further securing platform reliability.

Frequency of Insights Reports Publication

  • Report Publication Commitment: Throughout Season 2, the AWG aimed to deliver weekly insights reports to keep the community informed on trends, financial flows, and the overall health of the Inverse Finance protocol.
  • Achievement and Allocation: Out of 31 weeks in the season, 23 reports were successfully published. This represents a fulfillment rate of approximately 71% of the target, demonstrating a significant effort toward transparent communication with the DAO.
  • Resource Allocation Note: The remaining reports were not published due to strategic resource reallocations. Specifically, team efforts were redirected toward high-impact projects, such as the development of the Financial Machine Learning framework, Smart Contract Indexing, and the improved monitoring and alerting systems. These projects were prioritized to enhance the DAO’s data infrastructure and analytics capabilities, positioning the AWG for greater long-term impact.

Additional Projects for Season 2

  1. Advanced Data Connectors and Visualization Tools :

“Our objective to integrate more diverse data sources and improve our visualization tools has been achieved, we now aim to continuously improve them.”

Data Source Integration: The AWG successfully transitioned to dRPC, a decentralized data provider, enhancing the platform’s resilience and decentralization. This shift aligns with the DAO’s values and ensures a more robust data infrastructure.

User Experience Improvements: Several key features were added to improve usability and flexibility:

  • Dark Mode: Introduced a dark mode, enabling users to work more comfortably during extended hours.
  • Graph Customization: Implemented functionalities allowing users to save specific graph parameters, such as color coding, node size criteria, and positioning. Users can now edit graphs, delete or reposition nodes, and access saved configurations for consistency in ongoing analysis.

Support for DAO Working Groups: Visualizations are regularly provided to other DAO working groups, assisting them in interpreting data and making informed decisions.

  1. Machine Learning and AI Capabilities :

“We are planning to expand our application of machine learning and AI, aiming to build predictive models and conduct trend analysis that can anticipate market movements and inform strategic decisions.”

We are proud to introduce an advanced machine learning workflow into Inverse Watch, built on the best practices and principles from Advances in Financial Machine Learning by Marcos LĂłpez de Prado. This implementation reflects our commitment to leveraging cutting-edge financial ML methodologies, setting a new standard in data-driven decision-making and risk analysis within Inverse Finance DAO.

  • Integration with Inverse Watch: The AWG has embedded a flexible machine learning workflow directly within Inverse Watch, our fork of Redash. This integration leverages the platform’s data visualization and query functionalities, streamlining the process of preparing and feeding data into ML models. Users can now seamlessly pull data from multiple sources, perform preprocessing, and use it in models, which enhances our ability to make data-driven decisions.
  • Workflow Components: The ML workflow encompasses a comprehensive data pipeline:
    • Data Preparation: Queries fetch raw data, which undergoes cleaning, structuring, encoding, and scaling. Timestamp features are engineered, and autoencoders handle dimensionality reduction, optimizing data for model training.
    • Feature Engineering: Advanced techniques like cyclical encoding for time features and automated transformations streamline feature selection, while autoencoders manage complex data dimensionality.
    • Model Training and Tuning: Supported regressors—ranging from Linear Regression to Neural Networks—are selected based on specific configurations. Models are trained on partitioned datasets, with an auto mode enabling automated hyperparameter tuning.
    • Prediction and Deployment: Models generate predictions, translating them into human-readable outputs. Each model is serialized and stored, ensuring easy retrieval and deployment for ongoing use.
  • Supported Regressors: The system supports an array of regressors tailored for different use cases:
    • Linear/Logistic Regression: Offers simplicity for basic trend analysis.
    • Random Forest: Suitable for ensemble learning and complex pattern detection.
    • AdaBoost and Gradient Boosting: Ideal for refining predictions with sequential corrections.
    • Neural Networks: Handles non-linear relationships, providing deep learning capabilities for high-impact predictions.
  • Multi-output and Customizability: Custom estimators, such as the TunedMultiOutputEstimator for traditional ML models and the TuneableNNRegressor for neural networks, facilitate multi-output predictions, allowing for more comprehensive scenario modeling.
  • Auto Mode: A key feature, Auto Mode, enables automated hyperparameter tuning across regressors, enhancing performance with minimal manual intervention.

This machine learning infrastructure enables us to conduct detailed trend analysis and predictive modeling, empowering the DAO with forward-looking insights critical for strategic decision-making.

Link to the doc : Machine Learning | Inverse Watch Docs

  1. Bots and Automation Projects :

“We plan to further develop and enhance our suite of AI bots and infobots, improving community engagement and operational efficiency through advanced automation and AI capabilities.”

Community Engagement with InverseAlerts:

The InverseAlerts bot has been instrumental in fostering engagement on Twitter, posting 727 tweets throughout Season 2. This active presence has not only provided timely updates but also strengthened the DAO’s visibility and interaction on social media.

Enhanced Stability of Infobots:

In response to our commitment to operational efficiency, we have significantly improved the stability of our infobots on Discord. These bots play a crucial role in regularly fetching on-chain price data, ensuring accurate and up-to-date market monitoring for the Risk Department. This enhancement in bot reliability supports critical functions for real-time risk assessment and decision-making.

  1. Alert monitoring page :

“We plan to build a mechanism on inverse Watch that would allow team members to collaborate on and manage the alerts built and triggered by our system.”

User-Owned Alerts

In Season 2, the AWG introduced a user ownership model for alerts, allowing each user to create and manage their own alerts. This enhancement empowers DAO members to set personalized monitoring criteria, increasing the relevance and effectiveness of alerting within the platform.

Comprehensive Monitoring Page

A new alert monitoring page was developed, providing a centralized location to access the history and details of all alerts. This page enables users to review past alerts easily, with functionalities such as status indicators, timestamps, and query details. The screenshot below illustrates the layout, displaying essential metrics like total alerts, unique users, top destinations, and query history.

Enhanced Collaboration and Management

By enabling users to track and manage alerts within a dedicated interface, this monitoring page improves operational transparency and collaboration. Team members can now view alert trends, analyze frequently triggered queries, and adjust monitoring parameters in response to data insights, further supporting the DAO’s risk management efforts.

Annex II - Season 2 Expenses

Monthly Cost History

Cost table

May 2024 5-2024 Digital Ocean $ 475.20 $ 475.20
May 2024 5-2024 Quicknode $ 49.00 $ 524.20
June 2024 6-2024 Digital Ocean $ 475.43 $ 999.63
June 2024 6-2024 Quicknode $ 49.00 $ 1,048.63
July 2024 7-2024 Digital Ocean $ 475.20 $ 1,523.83
July 2024 7-2024 Quicknode $ 49.00 $ 1,572.83
August 2024 8-2024 Digital Ocean $ 475.20 $ 2,048.03
August 2024 8-2024 dRPC $ 200.00 $ 2,248.03
September 2024 9-2024 Digital Ocean $ 475.20 $ 2,723.23
September 2024 9-2024 Quicknode $ 49.00 $ 2,772.23
October 2024 10-2024 dRPC $ 163.84 $ 2,936.07
October 2024 10-2024 Digital Ocean $ 475.20 $ 3,411.27
October 2024 10-2024 Coingecko $ 154.80 $ 3,566.07
November 2024 11-2024 Digital Ocean $ 475.20 $ 4,041.27
1 Like