FLock.io is a London-based startup aiming to create the world’s first on-chain neural processor. Our initiative addresses the pressing challenge of reconciling the escalating demand for data with potential data breaches. Our platform facilitates data owners in training machine learning models without revealing their source data, and subsequently achieving consensus on the results in a transparent and decentralised fashion.
Central to our vision is the democratization of AI. We aim to establish a decentralized machine learning community on Taiko, enabling the collaborative development of advanced ML models.
Category
Surprise Us 🤫
Timeline
Start Date: September 15, 2023 (assumed)
15 October 2023: Official launch of FLock.io on the Taiko platform.
15 January 2024: Achieve 10,000 transactions on-chain, specifically targeting DeFi and Gaming use cases.
Projected Completion: January 15, 2024 (Based on a 4-month project duration).
Project Plan
Key features and technologies we leverage are as below:
No More CollectingYour Data
FLock.io's foundation is Federated Learning (FL). With FL, machine learning models are trained without sharing or transmitting source data, enhancing data security and minimizing the risks of misuse and breaches.
Transparency and Trust
FLock.io harnesses Federated Learning (FL) in a decentralized setup, enabling collaborative machine learning model training without data sharing. Underpinning FLock.io is a decentralized blockchain platform, which eradicates central vulnerabilities and provides transparent, tamper-proof records. To bolster this system's trustworthiness, FLock.io incorporates Zero-Knowledge Proofs (ZKPs) to validate computations without revealing the underlying data, ensuring the FL system remains both accurate and private.
Immutably Logged Transactions
Distributed Ledger Technology (DLT) underpins FLock.io's robust system. Every machine learning interaction, be it weight aggregation or model parameter updates, is immutably logged on the blockchain. This ensures a permanent, unalterable record, providing users and developers alike with a clear audit trail of all activities.
Rewards for Collaboration
At the heart of FLock.io lies a collaborative spirit. We believe that the best innovations spring from collective effort, and hence, we reward every contribution to our ecosystem.
FLock.io has designed a unique peer-to-peer (P2P) review and reward/slash mechanism powered by on-chain smart contracts. This mechanism not only incentivizes participants for genuine model submissions but also rewards vigilant reviews of model parameters. Through our cryptocurrency issuance, major contributors transform into ‘shareholders’, thus reinforcing the ethos of collaboration and shared success.
FLock.io has already laid down its foundational core technology and successfully launched its testnet. Our current focus is on refining our platform, making it even more user-friendly and expanding its reach to encourage more users to create or fine-tune models within our ecosystem. Designed with interoperability in mind, the FLock.io platform is poised for seamless integration with Taiko, ensuring effortless cross-chain governance and on-chain activities. This initiative is in sync with Taiko's mission of nurturing novel on-chain applications in the cryptocurrency realm.
Our roadmap includes launching at least one community-focused campaign monthly. These campaigns, primarily in the form of hackathons, are aimed at enhancing user engagement with FLock.io. Each campaign will be curated to resonate with specific crypto use cases, such as crypto LLM, to ensure relevance and heightened interest from our community members.
Project Impact
Secure and Customized AI Solutions: Users will be able to access a secure platform to finetune and inference AI models tailored to their needs, thereby ensuring data integrity while enabling specialized applications, as illustrated here - https://docs.flock.io/get-started/flock-llm-model
Community-Centric Ecosystem: Users will benefit from a structured community engagement, transitioning from AI enthusiasts to informed participants in both the General and Developer communities, facilitating deeper involvement in machine learning and web3 innovations.
Decentralization and Data Ownership: With FLock.io's on-chain ML environment, users will be empowered to create and deploy machine learning models without ceding data ownership to big tech, promoting privacy and autonomy in the web3 era.
Team Information
Key team members:
Jiahao Sun, CEO; Director of AI at Royal Bank of Canada; AI Research Fellow at Imperial College London; Graduated from Computer Science, Oxford
Shuoying Zhang, CSO, previously worked in consulting at BCG and led product DD/development at Advanced Blockchain; PPE, Oxford
Zehua Cheng, lead engineer, is a serial entrepreneur and previously exited an AI company to Tencent, PhD in AI, Oxford
Additionally, our team includes four PhDs in computer science from Oxford and Imperial, and four engineers with a combined experience of 20 years.
Point of Contact
tg: @limianf email:limian@flock.io
Previous Work
Our previous work has centered on building a community and reputation, alongside developing specific use cases. We initially targeted the developer and academic communities. Notably, our design paper won the best paper runner-up at the first AI x Blockchain workshop at this year's Neurips, a top-tier AI conference. As for use cases, we've focused on healthcare (personalized health model) and finance (anti-money laundering model), collaborating with leading institutions in both areas. With our core technology and testnet already in place, we are eager to contribute to the Taiko ecosystem
GitHub Repository :
Public (limited): https://github.com/FLock-io
Private code repos access by request
Demo: https://www.youtube.com/watch?v=lB69wUIgdus
Additional Information
FLock.io's platform is unique because it combines deep learning and on-chain consensus, which offers users both privacy and interoperability. Existing distributed learning solutions often struggle with throughput and model inversion attacks, but FLock.io's platform leverages the latest developments in high-performance computing and cryptography to guarantee performance and security against attacks. We also incentivize contributions from the wider developer and researcher community with the cryptocurrency they issue, and significant contributors become shareholders in the platform's success.
Please refer to https://flock.io/ for comprehensive information on our technology and documentation. We invite you to explore our testnet, which is currently accessible to everyone at no cost.
Agreement
[X] I agree to comply with the terms and conditions of the grants program
Project Description
FLock.io is a London-based startup aiming to create the world’s first on-chain neural processor. Our initiative addresses the pressing challenge of reconciling the escalating demand for data with potential data breaches. Our platform facilitates data owners in training machine learning models without revealing their source data, and subsequently achieving consensus on the results in a transparent and decentralised fashion.
Central to our vision is the democratization of AI. We aim to establish a decentralized machine learning community on Taiko, enabling the collaborative development of advanced ML models.
Category
Surprise Us 🤫
Timeline
Start Date: September 15, 2023 (assumed)
15 October 2023: Official launch of FLock.io on the Taiko platform.
15 January 2024: Achieve 10,000 transactions on-chain, specifically targeting DeFi and Gaming use cases.
Projected Completion: January 15, 2024 (Based on a 4-month project duration).
Project Plan
Key features and technologies we leverage are as below:
No More CollectingYour Data FLock.io's foundation is Federated Learning (FL). With FL, machine learning models are trained without sharing or transmitting source data, enhancing data security and minimizing the risks of misuse and breaches.
Transparency and Trust FLock.io harnesses Federated Learning (FL) in a decentralized setup, enabling collaborative machine learning model training without data sharing. Underpinning FLock.io is a decentralized blockchain platform, which eradicates central vulnerabilities and provides transparent, tamper-proof records. To bolster this system's trustworthiness, FLock.io incorporates Zero-Knowledge Proofs (ZKPs) to validate computations without revealing the underlying data, ensuring the FL system remains both accurate and private.
Immutably Logged Transactions Distributed Ledger Technology (DLT) underpins FLock.io's robust system. Every machine learning interaction, be it weight aggregation or model parameter updates, is immutably logged on the blockchain. This ensures a permanent, unalterable record, providing users and developers alike with a clear audit trail of all activities.
Rewards for Collaboration At the heart of FLock.io lies a collaborative spirit. We believe that the best innovations spring from collective effort, and hence, we reward every contribution to our ecosystem. FLock.io has designed a unique peer-to-peer (P2P) review and reward/slash mechanism powered by on-chain smart contracts. This mechanism not only incentivizes participants for genuine model submissions but also rewards vigilant reviews of model parameters. Through our cryptocurrency issuance, major contributors transform into ‘shareholders’, thus reinforcing the ethos of collaboration and shared success.
FLock.io has already laid down its foundational core technology and successfully launched its testnet. Our current focus is on refining our platform, making it even more user-friendly and expanding its reach to encourage more users to create or fine-tune models within our ecosystem. Designed with interoperability in mind, the FLock.io platform is poised for seamless integration with Taiko, ensuring effortless cross-chain governance and on-chain activities. This initiative is in sync with Taiko's mission of nurturing novel on-chain applications in the cryptocurrency realm.
Our roadmap includes launching at least one community-focused campaign monthly. These campaigns, primarily in the form of hackathons, are aimed at enhancing user engagement with FLock.io. Each campaign will be curated to resonate with specific crypto use cases, such as crypto LLM, to ensure relevance and heightened interest from our community members.
Project Impact
Secure and Customized AI Solutions: Users will be able to access a secure platform to finetune and inference AI models tailored to their needs, thereby ensuring data integrity while enabling specialized applications, as illustrated here - https://docs.flock.io/get-started/flock-llm-model
Community-Centric Ecosystem: Users will benefit from a structured community engagement, transitioning from AI enthusiasts to informed participants in both the General and Developer communities, facilitating deeper involvement in machine learning and web3 innovations.
Decentralization and Data Ownership: With FLock.io's on-chain ML environment, users will be empowered to create and deploy machine learning models without ceding data ownership to big tech, promoting privacy and autonomy in the web3 era.
Team Information
Key team members:
Jiahao Sun, CEO; Director of AI at Royal Bank of Canada; AI Research Fellow at Imperial College London; Graduated from Computer Science, Oxford
Shuoying Zhang, CSO, previously worked in consulting at BCG and led product DD/development at Advanced Blockchain; PPE, Oxford
Zehua Cheng, lead engineer, is a serial entrepreneur and previously exited an AI company to Tencent, PhD in AI, Oxford
Additionally, our team includes four PhDs in computer science from Oxford and Imperial, and four engineers with a combined experience of 20 years.
Point of Contact
tg: @limianf email:limian@flock.io
Previous Work
Our previous work has centered on building a community and reputation, alongside developing specific use cases. We initially targeted the developer and academic communities. Notably, our design paper won the best paper runner-up at the first AI x Blockchain workshop at this year's Neurips, a top-tier AI conference. As for use cases, we've focused on healthcare (personalized health model) and finance (anti-money laundering model), collaborating with leading institutions in both areas. With our core technology and testnet already in place, we are eager to contribute to the Taiko ecosystem GitHub Repository : Public (limited): https://github.com/FLock-io Private code repos access by request Demo: https://www.youtube.com/watch?v=lB69wUIgdus
Additional Information
FLock.io's platform is unique because it combines deep learning and on-chain consensus, which offers users both privacy and interoperability. Existing distributed learning solutions often struggle with throughput and model inversion attacks, but FLock.io's platform leverages the latest developments in high-performance computing and cryptography to guarantee performance and security against attacks. We also incentivize contributions from the wider developer and researcher community with the cryptocurrency they issue, and significant contributors become shareholders in the platform's success.
Please refer to https://flock.io/ for comprehensive information on our technology and documentation. We invite you to explore our testnet, which is currently accessible to everyone at no cost.
Agreement