Summary:
Delegates are challenging to choose and track, resulting in misaligned or inactive governance participation.
Project DAVOS is an AI-powered governance assistant that helps Polkadot users match with the right delegate based on shared values (Ethos) and delegate voting history.
We are requesting $25,000 (in 2 milestones over 6 weeks) to transform our mockup and validated concept into a working MVP, enabling delegate-user matching, user profile generation, and governance insights through a functional interface.

Problem:
Delegation in OpenGov is technically simple, but practically difficult. Users are faced with a list of delegate addresses and minimal context. They must manually research forum threads, track participation, and guess alignment on critical issues such as treasury management, technical upgrades, or community priorities. As a result, many token holders either delegate blindly, not at all, or to popular names without understanding their governance style. This contributes to low engagement and misaligned delegation.

What is Project DAVOS?
Project DAVOS (Decentralized Autonomous VOting System) is an AI-powered platform designed to enhance participation in Polkadot's OpenGov by making delegation smarter, more transparent and value-aligned. Users define their governance Ethos (either through templates or natural language), and DAVOS matches them to compatible delegates using AI-generated profiles derived from historical voting behavior. It also offers proposal Digests and reasoning Insights to reduce the time and complexity of staying informed.

This is the first iteration of DAVOS. The long-term goal is to build a fully reliable, auditable governance agent to support both token holders and delegates across ecosystems.
DAVOS is not designed to replace governance discussions: rather, it strengthens them. It improves matchmaking between users and delegates to:

• Helps users understand delegate values, through analysis and insights from past actions and votes
• Makes delegates’ behavior and values easier to understand, helping them reach aligned supporters
• Enhance understanding of OpenGov trends and alignment

We’ve already built a mockup, check it out!

Technical Design and Data Sources
DAVOS uses top publicly available LLMs (like ChatGPT, Claude, DeepSeek) with a Retrieval-Augmented Generation (RAG) approach. This enables the system to generate reasoning and insights based on verifiable, current information, without needing private model training or fine-tuning.
To reduce hallucinations and improve reliability, DAVOS combines outputs from multiple models. This cross-model strategy increases the factual accuracy of delegate profiles, summaries, and alignment scores.

Data sources for DAVOS include onchain voting history and governance discussions from key platforms like the Polkadot Forum, Polkassembly, and Subsquare. We’re also aware that valuable governance discussions happen on less structured channels. These will be addressed in future iterations.

Since the full dataset is vast, the MVP will limit the number of sources and delegates analyzed to ensure quality and traceability. The focus will be on building accurate delegate profiles and meaningful user-delegate matches.

The MVP will initially run on a centralized backend for simplicity and speed of development. Once the MVP work is completed, we plan to evolve DAVOS toward a more secure and decentralized architecture using Trusted Execution Environments (TEEs). TEEs will allow us to prompt LLMs from a secure enclave, ensuring that prompts, outputs, and reasoning (when local LLMs are used) cannot be manipulated and indeed match the open-source offchain code logic. This will make DAVOS’s behaviour fully traceable, auditable, and resistant to tampering. By leveraging a distributed network of TEEs, we can eliminate central points of failure and provide cryptographic guarantees over delegate recommendations. These safeguards will also help mitigate potential manipulation attempts, such as third parties trying to influence the matching algorithm. Our team has deep experience with TEE-based secure compute and is well-positioned to implement this in the next development phase.

Accuracy and traceability are key. Delegate behavior shifts will be flagged via Digest or notifications, and alignment is inferred from contextual reasoning, not static formulas, making the system harder to game and more trustworthy.

You can read more details in our attached document “Design and Data Processing Methodology”, available here.

---

Project DAVOS was first pitched during AAG 234 (link), and discussed here and in the Forum

We've validated the problem and solution through conversations and feedback within the Polkadot ecosystem and with other relevant DAOs in different ecosystems.

With this specific funding, we aim to develop and release a functional MVP that allows users to define their governance profile (Ethos), discover aligned delegates, and understand delegate behavior through summarized voting insights.

Milestones & Deliverables:

Milestone 1: Delegate + User Ethos Profiling MVP

• User Ethos, with the interface for selecting among different profiles or providing personal input through a prompt
• Delegate political profile and Ethos, based on on-chain voting history and information publicly available
• Matching algorithm, between the user and the delegate ethos
• Digest summary format (daily/monthly), with summary of the ongoing/recent discussion
• First UI prototype

Delivery timeline: 3 weeks from funding
Funding Requested: $15,000

---

Milestone 2: Expanded Matching & Governance Dashboard

• Enhanced configurability of the matching logic between the user and the delegate ethos
• Integration of active OpenGov referenda and delegates' votes
• UX improvements based on user feedback
• Finalize MVP, dashboard, and delegate profiles

Delivery timeline: 3 weeks after Milestone 1
Funding Requested: $10,000

Budget Breakdown:
This proposal covers the development of the DAVOS MVP. A future proposal will be submitted to support the production-ready version with expanded features.
Please find here the cost breakdown: Project DAVOS cost breakdown

Milestone	Category	Hourly Rate	Hours	FTEs	Total
1	Technical Work	$120	100	2	$12,000
1	Project Management	$120	25	1	$3,000
				Milestone 1:	$15,000
2	Technical Work	$120	68	2	$8,160
2	Project Management	$120	9	1	$1,080
2	Marketing	$100	10	1	$1,000
				Milestone 2:	$10,240
				Grand Total:	$25,240

Team:
Our team is complete and deeply experienced in smart contract development and DAO infrastructure. The team includes:

• Matteo – Business Developer: Focused on ecosystem design and DAO integrations. Former PM at pNetwork. Leads DAVOS’s governance strategy and community engagement.
• Thomas – Lead Protocol Engineer: A pioneer in decentralized finance and oracle systems, with a strong background in DAO tooling and governance infrastructure. Led the strategy and development of onchain voting, identity, and treasury management systems across multiple networks.
• Enrico – Full Stack & Protocol Developer: Frontend/backend contributor to AI UX flows, DAO dashboards, and privacy-preserving agent infrastructure using TEEs.
• Marco – Senior DevOps Engineer: Maintains secure infrastructure for blockchain services. Expert in Kubernetes, CI/CD, and blockchain node orchestration.

Call for Feedback:
We welcome community comments and suggestions to strengthen this proposal.
Please share your thoughts and questions on the scope, milestones, and funding.