Scaling Ethical, Decentralized AI

[karakrysthal]

Project Overview and Objectives

Mission

SI<3> (https://si3.space/) is a values-driven emerging tech ecosystem focused on fostering on-chain education, decentralization, and collaboration. One of our core facilitators of this will be Agent Kaia - an AI agent that facilitates meaningful connections and knowledge-sharing within SI<3> and the Web3 industry at large.

This proposal focuses on supporting growth opportunities for 0G.AI’s ecosystem through our Grow3dge Incubator, and integrating 0G infrastructure into Agent Kaia to enable verifiable identity, auditable matching, and portable reputation across platforms.

Team Background and Experience

SI<3> has been the recipient of nine grants from Web3 organizations including Arbitrum, Livepeer, Unlock Protocol, Push Protocol, Rari Foundation, Aeternity Foundation, Public Nouns, MoonDAO, and Rari DAO. We are also funding our organization through our training programs, including Grow3dge.

Here is a sampling of our testimonials:

“The Granting Access event was well-attended and successfully executed, with clear evidence of deliverables. The focus on onboarding under-represented groups is seen as a positive step that could drive significant impact for Arbitrum.” - ArbitrumDAO

“Push x SI<3> Ecosystem’s partnership has been impressive, driving awareness and hundreds of new adopters to the protocol.” - Push DAO

“SI<3> did a great job bringing people together in their Grow3dge program. I’ve been able to make meaningful connections with other Grow3dge members across DAO’s, Web3 companies, and local communities. I’ve already started applying some of the frameworks shared in the sessions to refine our existing programs, and I’m excited to keep learning and meeting more amazing people!” SzuTung Chen, DevRel Marketer at StarkNet Foundation

“I have experienced real value within the Grow3dge program, with significant partnership opportunities that have been made from integrations to event partners. I’ve also very much enjoyed the educational programming and learning from other industry experts. This is a program rich with value, and I’m excited to continue on the Grow3dge journey.” - Sohobit, Rari DAO

"Kara & SI<3> did an incredible job curating, producing and hosting an engaging, high-profile LinkedIn Live event for MoonDAO. The results speak for themselves: 500+ space professionals RSVP’d, and afterwards over a dozen business leaders reached out for partnerships and business opportunities.” - Pablo Moncada-Larrotiz, Founder & Executive Director of MoonDAO

Key Team Members

Kara Howard - SI<3> Co-Founder & Ecosystem Growth Lead

Kara is a full-stack Web3 growth, partnerships & community leader. For the last twelve years, she has been leading womxn-in-tech communities and developing community-driven growth programs and platforms for emerging technologies. Earlier in her career, Kara worked in equity research and investment banking and holds an MBA from NYU. Kara is a strong decentralization advocate and enjoys working at a systems level to architect scalable networks and programs that drive real-world Web3 adoption.

Jelena Gjorgev - SI<3> Co-Founder & CTO

Jelena Gjorgjev is a full-stack developer, blockchain strategist, and technical educator who operates at the intersection of technology, education, and real life. With a background spanning Web3, AI, data, and software engineering—and years of hands-on building, teaching, and leading—she is known for turning complex systems into practical, human-centered solutions.

Yasir Khan - SI<3> Technical Lead

Yasir is a full-stack developer and technology entrepreneur with over nine years of experience building scalable Web3 and blockchain products. He specialises in smart contracts, token systems, and decentralized application architecture, with a strong emphasis on production readiness and long-term scalability. He combines deep technical expertise with a startup-driven approach to product development, focusing on rapid execution, real-world impact, and long-term scalability.

Annie Brown, Reliabl.ai CEO
Award-winning AI researcher with over a decade of experience in technology startup leadership. Her work integrates machine learning, semiotics, ontology design, and feminist technology to build participatory, production-grade AI systems that perform better in real-world conditions. Annie has led AI bias research and model evaluation projects in collaboration with academic institutions, global partners, and frontier model developers, and has advised governments, foundations, and public-interest organizations on responsible AI deployment. Her work has been presented at ICML and NeurIPS, featured at the United Nations and the White House, and covered by Fast Company, Mashable, and TechCrunch.

Dr. Duncan McElfresh, Reliabl.ai CTO

Machine learning engineer and applied mathematician with deep expertise in model training, evaluation, and decision-making systems. He holds a PhD in Applied Mathematics and completed a postdoctoral fellowship at Stanford, where his research focused on optimization and ML applications in high-stakes settings. Duncan has built and deployed machine learning systems across industry and healthcare contexts, including decision-support tools used at Meta and Stanford Health.

Proposal Primary Objectives

1. Integrate 0G Infrastructure with Agent Kaia to enable verifiable, decentralized matchmaking records and cross-platform identity

2. Invite 10 0G Ecosystem Projects to our 6-month Grow3dge Builder Growth Incubator to support their growth foundations, marketing execution, and ecosystem collaboration, including ethical and technical AI support from Reliabl.ai

3. Build 0G.ai Integration connecting Agent Kaia with 0G.Ai for enhanced AI capabilities and cross-ecosystem knowledge sharing

4. Establish Verifiable Credentials System allowing users to prove SI<3> membership and match history on-chain

Impact Goals

• Decentralization and Public Accountability: Advance 0G Foundation’s mission to make AI accountable to the public by ensuring Agent Kaia’s matching logic is not only decentralized, but also transparent, verifiable, and aligned with human needs. This includes designing systems where communities can audit outcomes, understand why matches occur, and participate in improving the system over time.

• Trust & Transparency: Enable users to independently verify their match history and credentials

• Cross-Platform Identity: Create a seamless user experience across Telegram, Web, and future platforms

• Ecosystem Growth: Support ten 0G projects through our Growth Incubator program with the aim of achieving 2-3x+ user growth per active project in six months

• Decentralized Knowledge: Accelerate knowledge-sharing with 30+ companies, 45+ communities and hundreds of professionals within SI<3>'s ecosystem

• Ethical and Human-Centered AI: Enrich AI projects with ethical AI values and strategies to support a more optimal long-term outlook of decentralizing AI projects

---

Technical Architecture and Implementation Plan

Current Architecture

Agent Kaia is built on:

• Framework: ElizaOS (multi-agent framework)

• Runtime: Node.js 22, TypeScript 5.6.3

• Database: MongoDB 6.3.0 (primary), PostgreSQL 8.13.0 (supported)

• AI/ML: OpenAI GPT-4o-mini, OpenAI Embeddings

• Infrastructure: Docker, Fluence (decentralized cloud)

• Integrations: Telegram Bot API, REST API for web interface

Current Features:

• 15-step multilingual onboarding flow (EN/ES/PT/FR)

• Intelligent matching algorithm (weighted compatibility scoring)

• Cross-platform identity (email-based)

• Real-time match notifications

• Analytics and metrics tracking

• Role-based onboarding

• Web-based onboarding and matching experience (in addition to Telegram), enabling broader accessibility and improved user conversion

Proposed Solution

We will build a Conversational Matching Agent that:

1. Uses OG wallets as a decentralized identity (DID)

2. Stores encrypted user interest profiles on OG decentralized storage

3. Uses on-chain AI models to compute interest similarity and match confidence

4. Enforces matching rules and consent through OG smart contracts

5. Produces verifiable, auditable matching outcomes

Incentivised knowledge-sharing

Drip Rewards will be integrated as a community rewards and incentive mechanism to encourage knowledge sharing, thoughtful participation, and high-quality engagement around Web3 content.

Drip is used to:

• Reward users for valuable questions, answers, summaries, and insights

• Incentivize learning-first behavior, not spam

• Build a self-sustaining knowledge-driven community

• Create social proof and reputation for contributors

Knowledge is still created organically by the community — drip.re ensures contributors are recognized and rewarded.

AI inference is used selectively and deterministically, focusing on:

• Interest similarity scoring

• Reputation and quality classification

Key Features

On-Chain Interest Graph

Concept

• Each user has an on-chain profile hash

• Interests stored as:

  • Hashed tags

  • ZK-verified traits (privacy-preserving)

• Matching logic uses verifiable on-chain signals

OG L1 becomes the source of truth for identity + interests.

Decentralized Identity (DID) & Profiles

Integration

• Use OG L1 (EVM) wallets as identity

• One wallet = one social profile

• Portable across dApps

  Smart-Contract–Driven Matching Logic

  Instead of:

  • AI matches users privately

  AI proposes → smart contract verifies → user consents

  Benefits

  • Transparent matching rules

  • Auditable bias controls

  • On-chain fairness

    Technical Architecture (OG L1)

    Core Design Principles

    • AI inference is verifiable, not opaque

    • Storage is user-owned and portable

    • Smart contracts enforce fairness and consent

    • Sensitive data remains encrypted and off-chain

      

OG Foundation Technology Usage

OG Capability Usage
On-chain AI Interest similarity & reputation inference
Decentralized Storage Encrypted user profiles & embeddings

Smart Contracts Matching, consent, incentives
Wallet Identity DID & access control
Event Logs Social analytics & transparency

Privacy & Ethical Design

• Raw conversations are never stored on-chain

• User profiles are encrypted before storage

• Only hashes and AI outputs are committed to OG L1

• Users explicitly consent before any match

• Reputation is non-transferable (soulbound)

This ensures compliance with privacy-by-design principles.

Proposal Timeline & Deliverables: Technical Milestones

The technical aspect of this proposal focuses on integrating 0G’s decentralized identity, storage, and verifiable compute capabilities into Agent Kaia, enabling deterministic, auditable, and consent-driven matching. Across three phases, we will deliver a working end-to-end system that demonstrates how on-chain AI can support real-world social coordination while preserving privacy, user agency, and trust.

Phase 1: Identity & Profile Infrastructure

Timeline: Weeks 1-4

Objective: Establish decentralized identity and user-owned profile storage using OG L1 primitives.

Technical Scope

• Integrate OG wallet-based sign-in as the primary user identity

• Generate a Decentralized Identifier (DID) per wallet

• Define interest schema (tags, vectors, metadata)

• Encrypt user profiles client-side

• Store encrypted profiles on OG Decentralized Storage

• Commit content hashes to OG L1 smart contract

• Implement consent signature flow for profile updates

Key OG Integrations

• OG Wallet / DID

• OG Decentralized Storage

• OG L1 smart contracts (profile registry)

Deliverables

• Working wallet login flow

• On-chain DID + profile hash registry

• Decentralized encrypted profile storage

• Technical documentation for profile schema

Success Criteria

• Users can create, update, and retrieve profiles using OG wallet

• Profile data is fully user-owned and portable

• No centralized database dependency

Phase 2: On-Chain AI Matching & Verification

Timeline: Weeks 5–8
Objective: Demonstrate verifiable interest-based matching using OG’s on-chain AI models.

Technical Scope

• Convert user interests into embeddings (off-chain preprocessing)

• Pass embeddings or hashed vectors to OG on-chain AI model

• Implement deterministic similarity scoring

• Define match confidence thresholds

• Smart contract verifies AI output before match approval

• Users explicitly consent to matched conversations

• Log AI inference results on-chain for auditability

Key OG Integrations

• OG On-Chain AI Runtime

• OG Smart Contracts

• OG Event Logs

Deliverables

• Deployed on-chain AI inference pipeline

• Match verification smart contract

• Transparent, auditable match scoring

• Demo of AI-verified matching flow

Success Criteria

• AI-based match scores are reproducible & verifiable

• Matching logic is transparent and auditable

• No opaque, centralized matching decisions

Phase 3: Reputation, Incentives & Demo Release

Timeline: Weeks 9–12

Objective: Add trust, reputation, and incentives; release public demo and documentation.

Technical Scope

• Implement reputation scoring logic (non-transferable / soulbound)

• Reputation updated based on:

  • Match success

  • Conversation feedback

  • On-chain AI classification

• Incentivize positive participation using OG tokens

• Add abuse/spam detection signals

• Build end-to-end demo UI

• Publish open documentation and reference architecture

Key OG Integrations

• OG Smart Contracts (reputation & incentives)

• OG Tokens (rewards / penalties)

• OG Event Analytics

Deliverables

• Reputation & incentive contracts

• End-to-end working demo (Web / Mobile)

• Public technical documentation

• Final report to OG Foundation

Success Criteria

• Reputation discourages spam and bad actors

• Incentives drive high-quality interactions

• Demo showcases full OG L1 stack usage

Phase 4: Launching internal beta, testing and feedback review

Timeline: Weeks 13–24

Objective: Add trust, reputation, and incentives; release public demo, test and develop documentation.

---

Technical Financial Milestones:

1. Technical project kickoff (Month 1): $5k USD

2. Mid-project completion (Month 3): $5k USD

3. Project completion (Month 6): $5k USD

Proposal Timeline & Deliverables: Business Milestones

The business aspect of our proposal includes inviting 10 promising 0G.ai-supported project to attend our six-month Grow3dge Builder Growth Incubator. Our AI agent will also be introduced to the attendees of Grow3dge. This is a sub-program within our broader initiative, in which all participants receive access: SI3 Ecosystem

Grow3dge aims to help solve the following problems:

• Builders struggle to grow their projects after initial MVP as they lack marketing experience

• Blockchains and protocols are struggling to see ROI on their grant funding due to lack of project adoption

• Small DevRel teams have limited bandwith and resources to support their builders in their growth at the level they require

• Builders experience a lack in connectivity with other builders and are often creating in silos

Our solution:

• Grow3dge’s Builder Growth Incubator offers a structured 6-month framework and launchpad with growth-related education and activities, expert mentorship, and hands-on support

• DevRels and their builders participate in a collaborative learning environment in Ro.am to bolster ongoing education, collaborative learning, and receive strategic marketing support throughout the program

• Thoughout the incubator, ecosystem and their DevRels receive strong insights into which builders are dedicated and performing, which provides a deeper understanding into which projects may be a fit for future grants and resource allocation

• Builder teams are able to connect with other teams and protocols for improved collaborative growth opportunities

Incubator KPI’s:

These are the qualitative and quantitative metrics that we aim to reach with each participating ecosystem:

• 80% builder team completion rate, which entails builder teams actively participating and submitting 80% of the action steps outlined in the programming

• 70-80% of teams doubling their user growth or more from the incubator

• Ecosystem and devrel teams have 2-3x greater understanding of their high-potential builders and their products and growth rates throughout and post-incubator

  Improved ecosystem cohesion and collaboration between builders seen through partnerships and integrations with the ecosystem builders as well as cross-chain collaborations

• Loyalty and referrals increased within the builder network due to the support offered to them in this Grow3dge incubator, acknowledging that their ecosystem supports the growth of their projects beyond just technical building

Milestone 1: Grow3dge Builder Growth Incubator Foundations

Timeline: Months 1-3

Objective: Identify and onboard ten promising 0G.AI Foundation projects into our incubator. Projects participate in our ‘Marketing Foundations’ and ‘Know Your Customer’ modules.

Key Participant Activities & Deliverables:

SI<3> manages the 10 builder teams in delivering these tasks with weekly strategic reviews, and co-working and advisory availability within our Roam virtual coworking space

• Watch our ‘Marketing Foundations’ video to prepare analytics, KPI’s, channels and marketing funnels.

• Set up and/or review your marketing analytics dashboard (custom dashboard template is provided with instructions - requires some technical knowledge) and start tracking your metrics weekly.

• Select 2-3 channels & KPI’s to focus on (email, social, user growth) and focus on executing channel growth throughout the program, and exploratory growth strategies.

• Map your growth marketing funnel and track your metrics at each stage.

• Watch our ‘Know Your Customers’ video to establish clarity over your customers, their pain points, and early community-market fit assessment

• Receive SI<3>'s Persona Mapping Canva template and map your customer personas

• Hold 3-4 existing and/or new potential customer interviews, receiving insights to help shape your product and market fit. Summarize your users’ pain points and opportunities from your metrics analysis and customer interviews, and develop the problem and solution statement that your organization is solving.

• Assess early community-market fit stages (receive our C/M/F template in Fibery)

Milestone 2: Go-to-Market & Grow

Timeline: Months 3-6
Objective: Support builder teams in their marketing & communications strategies and activities, and foster collaboration and partnerships between the ten participating builder teams and our other Grow3dge program participants.

Key Participant Activities & Deliverables:

SI<3> manages the 10 builder teams in delivering these tasks with weekly strategic reviews, and co-working and advisory availability within our Roam virtual coworking space

• Watch our ‘Go-to-Market’ video to start taking action towards your marketing, business development & content activities

• Develop your marketing campaign outline and content development needs, and map your campaign launch timeline (receive our campaign launch template in Fibery), including the opportunity to launch your content within SI<3>'s free and open Scholars tier in our social learning platform.

• Initiate marketing content development

• Continue development of your marketing campaign, business development partners, and content

• Watch our ‘Community Growth’ video to learn how develop, manage and incentivize your communities

• Develop your community, content and incentivization strategy

• Begin to execute your community strategy and track your metrics

• Launch your marketing campaign through your community channels! (and track your metrics) and track C/M/F (Community Market Fit)

• Watch our ‘PR & Partnerships’ video to learn how to develop relationships with journalists and communities for paid and organic PR, and how to develop meaningful partnerships and negotation skills

• Map out 2-3 partnership & referral opportunities and begin to take action towards those, with partnership agreements

• Follow and engage with journalists and develop your press kit (receive our CRM template in Fibery) and develop your PR announcement

• Announce your product with your partners and potential journalists

• Watch our ‘Fund Your Growth’ video to learn how to prepare your project and pitch to share with investors and ecosystems for funding

• Review the performance (with Grow3dge team) of your go-to-market activities and rate your channel growth by highest performing to least performing. Select 1-2 channels that are not activated that you would like to test. Set up and test one new growth channel, and build a roadmap for continued testing. Also outline a marketing roadmap (recommend testing one new strategy per month). Forecast how this channel could perform with the potential for paid incentives

• Prepare your pitch deck and practice pitching

• Prepare your financial model and practice speaking to your financial metrics

SI<3> will provide 0G a summary of learnings and reporting post-incubator with these metrics, and a survey summary from builders and their experiences.

Ethical and Technical AI Advisory with Reliabl.ai

Reliabl will support SI<3> through ethical and technical advisory and implementation, with a focus on ensuring that Kaia’s on-chain AI matching is deterministic, verifiable, auditable, and community-aligned.

While 0G provides cryptographic and computational verifiability, Reliabl adds a human-centered verification layer that strengthens real-world trust, usability, and long-term adoption. This is especially important because verifiable matching only becomes meaningful to users when the embeddings, taxonomies, thresholds, and evaluation logic reflect the diversity of the community using the system.

Reliabl’s Support Includes:

• Inclusive Taxonomy and Schema Design: Reviewing and designing inclusive skill, interest, and opportunity taxonomies before they are embedded, hashed, or used in similarity computation. This ensures Kaia’s matching system can represent diverse professional contexts, non-Western opportunity structures, and emerging roles.

• Embedding Input Evaluation and Bias Stress-Testing: Evaluating embedding inputs and feature representations for representational bias, including stress-testing for failure modes that could distort outcomes for underrepresented builders, womxn, non-binary participants, and emerging market contributors.

• Fairness-Aware Similarity Scoring and Threshold Design: Supporting the definition of similarity scoring logic and match confidence thresholds that are transparent, explainable, and fairness-aware, reducing the risk of exclusionary or overly narrow matching behavior.

• Structured Evaluation and Annotation Rounds: Running structured evaluation rounds to validate match quality, including annotation and feedback loops that test whether the system produces matches that are relevant, equitable, and economically meaningful for different user groups.

• Transparency and Contestability Mechanisms: Supporting user-facing transparency and contestability so participants can understand why matches occur, provide feedback, and challenge outcomes, increasing trust and improving system quality over time.

Reliabl will be supporting both SI<3> and the 0G builder cohort, including two implementation workshops plus a hands-on technical and evaluation sprint during Phase 2 to directly support teams deploying on-chain AI systems.

Reliabl’s Work samples:

Skincare Product Recommendation AI with Osaz.ai

• Advised on bias reduction strategy and data architecture for AI-driven skincare recommendations affecting melanin-rich skin tones.

• Guided the replacement of third-party labeling with user-labeled, community-informed annotation systems.

• Provided technical guidance on taxonomy design, validation workflows, and feedback loops to improve recommendation accuracy.

• Outcome: More representative datasets and improved precision for underrepresented users.

Social Media Content Recommendation Systems with Communia

• Advised on ethical and technical design of content recommendation systems for a women-centered social platform.

• Provided technical guidance on user-driven taxonomy development, contextual annotation, and model fine-tuning strategies.

• Consulted on evaluation criteria and success metrics aligned with emotional safety and inclusivity.

• Outcome: Built proprietary recommendation system and improved relevance across diverse lived experiences.

Contextual Emotion Detection for Automotive AI with ToumAI

• Advised on emotion recognition system design for in-vehicle AI, emphasizing cultural and contextual accuracy.

• Guided the development of culturally grounded emotion taxonomies and annotation protocols for Japanese audio data.

• Provided technical oversight of QA processes, annotator training, and data validation.

• Outcome: improved emotion detection accuracy and reduced cultural bias in automotive deployments.

Embedding End Users in AI Model Evaluation with Humane Intelligence

• Advised on technical evaluation frameworks for assessing fairness, reliability, and human rights impacts of deployed AI systems.

• Co-designed structured red-teaming and Bias Bounty methodologies, embedding end-user perspectives into ground truth comparison.

• Provided guidance on audit workflows, failure-mode discovery, and evaluation reporting.

• Outcome: identification of systemic bias missed by standard benchmarks.

Business Financial Milestones:

1. Grow3dge Incubator kickoff (Month 1): $10k

2. Incubator Month 3 completion with 10 OG.AI mid-Incubator summary update (Month 3): $7.5k

3. Incubator completion with final summary report of 10 OG.AI projects and OKR’s acheived, including Reliabl’s technical and ethical AI support (Month 6): $7.5k