Blockchain projects have witnessed exponential growth during the last bull cycle (June 2021 to May 2022). DeFi, Gaming, and Metaverse projects attracted the interest of veterans and newcomers alike. However, a handful of ventures in this space integrate Artificial Intelligence (AI) and Machine Learning (ML) technological advancements with blockchain platforms. The applications are all-encompassing, including cybersecurity, research, and DeFi. Today’s global AI market stands at around $327.5 billion (2021).

Kosen Labs (ref. 1) aims to bring more efficiency and value to decentralized ecosystems by introducing the benefits of advanced Artificial Intelligence (AI) and Machine Learning (ML) models. As a first step, the project team attempts to build a solution to the problem of AMM (Automated Market Makers) Liquidity Providing through the Kosen Protocol (Kosen Labs v1 strategy) that could address these shortcomings and offer unique features to its users. 

The core of Kosen Protocol’s technological solution is the creation of asset vaults that enable Liquidity Providers (LPs) to maximize their Average Percentage Yield (APY). Essentially, Kosen vaults are machine learning-based automated liquidity management solutions. At the outset, these vaults will be deployed on Uniswap V3, followed by SushiSwap Trident. Over time, they intend to diversify the number and complexity of these pools.

There are two central parts to the protocol. Those are the vault strategy and the Keepers — providers of AI/ML strategies. The network’s native token, $KOSN, provides several utilities to the holders and will be the central governance element in the planned DAO. The token’s value will be directly linked to the Total Value Locked (TVL) in the protocol and the success of the keeper strategies because the protocol will apply a dynamic protocol fee.

The Project is located in Portugal, one of the friendliest blockchain and crypto jurisdictions. Two highly-skilled scientists and experienced professionals lead the project.

Kosen Labs is currently at the concept stage and has yet to launch any product. It does not have a clear roadmap but has already secured $5 million in a seed round from Andreessen Horowitz.

Our researchers gave Kosen Labs a final rating of C. The rating scale is available at the end of this report.



Introduction to Kosen Labs

Kosen Labs seeks to bring the benefits of advanced AI/ML and optimizations to the decentralized economy. The founding team envisions a research-first approach at the intersection of AI/ML and Web 3.0. In this journey, the first solution — the Kosen Labs v1 strategy or Kosen Protocol — is a Liquidity Provider (LP) yield optimization tool. Subsequently, Kosen aspires to deliver solutions for Maximum Extractable Value (MEV) and Yield Farming. A core element of the project will be the in-house incubator at Kosen Foundation, with a strong focus on research activities by attracting the best talent in blockchain and Web 3.0 development.

Kosen Labs Mission: “The mission of Kosen Labs is to bring the benefits of AI & advanced optimization to the DeFi space and beyond, all while keeping entry barriers low for the end-user and allowing everyone to partake in this new economy.”


Kosen Labs: Proposed Solution (there could be other solutions not included in this graph). Illustrated by D-Core.


Kosen Protocol


The blockchain ecosystem is still in its early days regarding the integration of machine learning and statistical optimization technologies. Kosen plans to deploy these techniques, which can act as a significant value multiplier for its users.

Kosen Protocol is the first product that Kosen Labs conceptualized and presented to the investors as an intuitively understandable solution where the founders could display the challenges and the benefits of applying ML in actively managing liquidity (ref. 2).

Essentially, the idea of the protocol is to be an ML-powered Liquidity Pool (LP) manager on automated market makers (AMMs), for example, Uniswap v3 and SushiSwap Trident. However, there is currently no Minimum Viable Product (MVP) available for users to test the solution. The project’s whitepaper (last updated nine months ago) explains the team’s short-term and long-term plans.

Given the unavailability of a detailed project roadmap, it is challenging to determine the development progress made by the team. Since the close of the seed funding round in November 2021, there hasn’t been any progress update available publicly. Presumably, the project is either stagnating or building in stealth.

Success Factors

Based on our understanding, there are several success factors for the project. These factors are listed below:

  • Addressing a potentially large and untapped market segment.
  • Its Founders’ experience in the AI/ML field.
  • Being backed by large and well-connected investment firms.



Market Conditions

Artificial Intelligence (AI) running on top of a blockchain can significantly enhance machine learning capabilities. Nonetheless, AI in blockchain or Web 3.0 context is a relatively under-explored area, or the existing solutions have not managed to gain noticeable traction (ref. 3). ARK Invest’s Big Ideas 2022 Report shows that five innovation platforms, including Artificial Intelligence and blockchain technology, are evolving and converging (ref. 4). The global AI market was estimated to be valued at around $327.5 billion in 2021 (ref. 5).

The AI sector in the crypto industry currently accounts for about $708 .4 million as of 29 June 2022 (ref. 6). This sector represents projects broadly connected to the AI industry in some way.

In a presentation (ref. 7) at MIT in 2020, Jamiel Sheikh, CEO of Chainhaus (ref. 8), explained that the blockchain-AI convergence process will cover four stages:

  • Stage I: Blockchain proof of concepts.
  • Stage II: Tokenization of assets on the blockchain.
  • Stage III: Digital investment assets traded on blockchains, powered by machine learning.
  • Stage IV: AI as economic agents that trade digital investment assets.

If successfully developed, Kosen Labs’ solution and products could be among the first to be placed on Stage III of the process mentioned above. The most significant use case for AI Information Technology and Telecommunications is cybersecurity. However, over the last two years, on-chain activity has increased tremendously on public blockchains, thus fulfilling the chief precondition for maturing AI technology and financial products on these networks. Above and beyond, 34% of financial services firms use Artificial Intelligence in operations. 

Similar to the traditional banking system, AI can improve the utility of DeFi services by making possible, among others (ref. 9): 

  • Smarter oracles.
  • Network fees optimization.
  • Limit orders.
  • User scoring.
  • Risk mitigation.
  • Automated hedging. 
  • Yield optimization.

DeFi went through a crucial change with the creation and rise of yield farming products as a central block in its realm. Before the DeFi Summer of 2020, only $1.03 billion of value was locked in DeFi protocols (as of May 25). This sum grew to a surprising $20.49 billion by the end of the year (ref. 10). Before the crash of May 2022, nearly $224 billion of value were locked in hundreds of DeFi protocols, of which the Total Value Locked (TVL) of Tricrypto (ref. 11) (USDT, wBTC, and wETH) was over $1.6 billion on Curve (CRV), nearly $220k on Arbitrum and over $50 million on Polygon (MATIC). Currently, the TVL of DeFi stands at $74.72 billion (ref. 12).


Total Value Locked in DeFi. Source: DeFi Llama (As of 29 June 2022). 

In 2021, both Uniswap v3 and SushiSwap deployed some version of a Layer 2 solution. With Ethereum completing its PoS transition (ref. 13) in Fall 2022, the Kosen Labs team expects AMM transaction gas fees to go down significantly. This reduction of fees would initiate a whole new spectrum of potential LP strategies and create an environment where successful LPs would be forced to use some form of advanced ML & optimization techniques to remain competitive. Alternatively, this would be the AMM version of market makers moving towards effective algorithmic/high-frequency trading strategies. This is the thesis upheld by Kosen in their whitepaper.


As discussed above, the crypto AI sector represents projects connected to the AI industry in several ways. For instance:

  • SingularityNET aims to decentralize the creation and application of AI to progress toward Artificial General Intelligence ultimately (ref. 14).
  • Cortex is a decentralized world computer capable of running AI and AI-powered dApps on the blockchain (ref. 15).
  • The Web 3.0 data platform Ocean Protocol tries to unlock data for AI (ref. 16).
  • Velas, for instance, set out to build an AI-powered dPoS, although it later became a Solana fork.
  • The Sentinel Protocol by Uppsala Security discussed in its 2018 Whitepaper that it aimed to use blockchain and AI in two different ways: to develop a machine learning-based blockchain security client wallet and construct a Fraud Detection System (FDS) using the data from the blockchain.
  • Numerai is a tournament where participants build machine learning models on financial data to predict the stock market (ref. 17).
  • Some projects that have now become dysfunctional or stagnant. E.g., Artificial Intelligence.

Kosen Labs is not the only project working at the intersection of AI and blockchain. Nonetheless, it is challenging to establish closer competitors to Kosen Labs.

Interestingly, since 2019, the giant Microsoft has been investing in AI products that use public blockchains. The company’s goal is to make products through the collaborative effort of the AI community, since the market is still dominated by applications that require access to private data. In this sense, Microsoft AI’s deliverables would enable customers to use machine learning for free and create products such as virtual assistants or recommender systems.

On the other hand, Kosen Protocol can be considered to compete with the broader Yield Aggregator category. This category, as per Defilama, consists of 54 protocols holding about $1.29 billion in TVL. However, most of these aggregators rely on algorithmic solutions to generate better APR for users. Interestingly, there are several examples of AI implementations. Out of these solutions, Oraichain comes very close to Kosen Protocol and yAI.Finance built on Oraichain is an AI-based Yield Aggregator (ref. 19). OptyFi, another close peer to Kosen Protocol, is also marketed as an AI-powered protocol to launch DeFi yield strategies (ref. 20). KingDefi touted itself as the first AI multichain yield aggregator, yet it hasn’t been active (ref. 21).


Current ROI (based on deposit start value) vs. competitors (August 2021). Source: Kosen Labs Whitepaper 


How is the project different from its competitors? 


As explained in its whitepaper, Kosen will focus on a decentralized ecosystem where it can add the most value. It has three criteria for assessment:

  • Complexity and tractability of the problem.
  • Current and future usefulness of the solution.
  • Viability of the implementation.

The three points mentioned above are not explained in detail in the whitepaper. The project reasons why DeFi, particularly active liquidity provision, provides the best fit to the above three elements:

  • When an industry is programmable, it is (hyper-)optimizable.
  • Next-gen AMMs offer a superior experience to end-users at the expense of LP complexity.
  • Next-gen AMMs are a perfect setting for sophisticated optimization techniques.
  • Ethereum and AMMs Layer 2 scaling will move LPs towards more intelligent strategies.




The first product planned by Kosen Labs is its AI-powered LP optimization solution on AMMs (initially on Uniswap V3 and then SushiSwap Trident). With no public open-source codebase or a minimum viable product for the Kosen Protocol, users have no opportunity to experience the product yet.

However, the project’s whitepaper briefly explains the concept behind the Kosen Protocol. Kosen asset vaults are an intelligent yield optimization layer deployed on AMMs. These vaults manage user LP positions autonomously to generate superior APYs. The asset vaults are run based on the ML-driven strategies contributed by Keepers. Kosen will be the first Keeper in the protocol. As discussed in the whitepaper, Kosen has built a proprietary AI research stack that will be used for the V1 strategy.


System Diagram for the Kosen Protocol. Source: Kosen Labs Whitepaper 


Kosen Protocol will be deployed initially on Uniswap V3 and SushiSwap Trident. Later in the development cycle, it is hoped that the protocol will be launched on other alternative networks.



Kosen Labs has not presented a detailed and time-bound roadmap yet. The whitepaper has briefly discussed its short-term and long-term plans.



Kosen Labs is led by two founders, Peter Toth and Miljan Martic, ex-Google DeepMind researchers and engineers. The project models they have previously worked on were used to defeat world champions in the game of Go, beat top professionals in StarCraft, and optimize billions of dollars in revenue across various Google products. Peter and Miljan are first-time founders.

Miljan Martic (ref. 22), Co-Founder and CEO, was a Senior Research Engineer at DeepMind and a Data Scientist at MarketInvoice and LoopMe. Earlier in his career, he was Research Associate at Signal Media Ltd. and Summer Analyst at Credit Suisse. Miljan worked as a developer at the University of Manchester’s Advanced Processor Technologies Research Group. He obtained his MSc in Machine Learning from UCL.

Peter Toth (ref. 23), Co-Founder and CTO, worked at Google DeepMind as a Research Engineer. At DeepMind, he co-authored Hamiltonian Generative Networks ICLR 2020. Peter held Data Science roles with several organizations, e.g., Agoda, Zalando SE, and Webtrekk. He studied at Utrecht University and worked as a Lawyer Trainee at Réti, Antall, Várszegi & Partners Law Firm.

Guven Toprakkiran (ref. 24), Engineering Lead, was a Lead Data Engineer at SoundCloud. He also held Data Engineering roles with several other entities.

Gurpreet Singh (ref. 25), COO, was the VP of Finance and People at Disperse. She is an Advisor to Imagine Zero. She obtained her BA and MSc in Accounting from Arizona State University.


There is no publicly available information on their advisors.

General Comments on the Team & Advisors

Both founders have excellent backgrounds and experience in data science.

The project recruits Blockchain Engineers, ML Engineers, and ML Research Engineers. These openings were posted three months ago (ref. 26).

Throughout our research, we did not find evidence that the team members have taken part in any previous or current illegal projects or projects that were controversial.




Kosen Labs is a Cayman Islands registered Corporation, as the trademark registration details disclosed (ref. 27).

In the long term, the protocol’s governance will be transferred to decentralized governance mechanisms via Kosan DAO. To lay the foundation for progressive decentralization, the Kosen Foundation will be formed. Kosen Labs will be created as part of the Kosen Foundation.


Kosen Labs completed a seed funding round led by Andreessen Horowitz. Several other investors participated in this funding round: Framework Ventures, Protocol Labs, Daedalus, and Formless Capital (ref. 28).


Kosen Labs Partners. Source: Kosen Labs

Legal Advisors

The project has not announced any legal advisors. However, a team member oversees all operations-related functions across Kosen Labs, including Legal, Finance, People Ops, and Recruiting.

Anastasija Verbicka (ref. 29), Operations Manager, was with Koinly immediately before joining Kosen Labs. She has also worked for Axora, Sintec UK Limited, Homie.rent, and Ofixu.com. She holds a Master of Laws (LLM) from the Queen Mary University of London and a BA in International Politics and Law from Middlesex University.


The project is not conducting a token offering yet.

Token Classification

The KOSN token is a utility token for the Kosen Protocol. The main utility of the KOSN token will be protocol governance.



Kosen Labs demonstrates a comparatively poor social media presence at present. An observation is that the team lacks marketing and content skills. In February 2022, the project announced its Agent Program (ref. 30), an effort to build a community around the project. Under the program, the project sought to enroll Regional Community Managers, Social Media Agents, Content Creators, and others. Understandably, the Program has not delivered effective results so far.

The project’s Twitter handle has been inactive for approximately two months as of this report. It has 4.8k followers.

It is noticed that the project’s Discord channel has recently seen a minor decrease in members to 5.5k. It has not been adequately developed into a platform for useful discussions about the project.

Kosen has not posted on LinkedIn for several months. The LinkedIn page has 1.0k followers.

Kosen Labs does not have its own YouTube channel, but several third-party channels have covered the project.

The Medium blog currently only hosts two articles.



The following list of risks is not an exhaustive one. Some of these risks are internal, and others are external to the organization. Some risks may be minor/not materialize:

  • Operational Risks: The project may face risks due to various unexpected operational issues. Some of such issues could be the reliance on the two key founders and the inability to source the required blockchain and AI/ML skills.
  • Lack of community support: If the project continuously fails to attract strong community support, that could lead to failure or stagnation of the project.
  • Market conditions: Extended periods with the crypto market underperforming could be a risk to the project. This risk could mainly manifest itself in difficulty in raising development funding.



Everything you see in this report is the aggregate result of an extensive research process carried out by a distributed team of researchers and crypto enthusiasts around the world. The process consists of 60 questions divided into three phases. Researchers are called to answer these questions about a project, while providing links or screenshots as evidence to support their answers. For every answer, they also provide a rating from zero to ten. The average of their ratings is detailed below.


Our researchers gave Kosen Labs a final rating of C.




This Report is for informational purposes only and/or all or any of its content thereof, should not, may not and will not be taken to constitute, either as a whole or in part, any investment advice or recommendation or similar, regulated, or authorized advice, and D-Core by producing, disseminating, giving away, or making available this Report does not, should not, may not and will not be taken to advise on investments, or carry out any similar activity, or any regulated activity or any other authorized activity. D-Core is not authorized by the Financial Conduct Authority or by any other competent EU or elsewhere or otherwise competent authority to carry out any regulated activities and/or any activities within the scope of these authorities’ competence. 

D-Core excludes and disclaims all liability and/or responsibility whatsoever and/or howsoever caused, arising out of any actions, or omissions taken, or made by any authorized and/or other recipient of this Report in reliance on, or arising out of, or in connection with any or all content of this Report. Any authorized and/or other recipient of this Report acknowledges, accepts and agrees that they carry out their own independent research and act in their own sole risk in reading or using any or all information contained in this Report. In any event, recipients of this Report are urged to seek professional advice before making any potential investment decision in relation to the project described herein. Any authorized and/or other recipient of this Report accepts this Disclaimer in full. For the avoidance of doubt, this Disclaimer is binding against any recipient of this Report whatsoever.



  1. “Kosen Labs.” https://www.kosen.ai/. Accessed 29 Jun. 2022.
  2. “Dove Dispatch #31 (Miljan Martic – Kosen Labs).” 14 Dec. 2021, https://dovemountain.substack.com/p/dove-dispatch-31-miljan-martic-kosen. Accessed 29 Jun. 2022.
  3. For further details, see the article by Raj Shroff, available at https://medium.com/swlh/when-blockchain-meets-artificial-intelligence-e448968d0482.
  4. “Big Ideas Report: Innovation Research by ARK Invest.” https://ark-invest.com/big-ideas-2022/. Accessed 29 Jun. 2022.
  5. “Artificial Intelligence (AI) worldwide – Statistics & Facts – Statista.” 16 Mar. 2022, https://www.statista.com/topics/3104/artificial-intelligence-ai-worldwide/. Accessed 29 Jun. 2022.
  6. “AI Coins – CryptoSlate.” https://cryptoslate.com/cryptos/ai/. Accessed 29 Jun. 2022.
  7. Link: http://bit.ly/DIAFrontiers
  8. “Chainhaus – new finance for a new future.” https://www.chainhaus.com/. Accessed 29 Jun. 2022.
  9. “Artificial Intelligence in Decentralized Finance – TrendingTopics.” 21 Sep. 2020, https://www.trendingtopics.eu/artificial-intelligence-in-decentralized-finance/. Accessed 29 Jun. 2022.
  10. “Survey of current nature of LPing in DeFi | by Instrumental Finance.” 9 Feb. 2022, https://medium.com/@instrumentalfinance/survey-of-current-nature-of-lping-in-defi-e9a5b9796393. Accessed 29 Jun. 2022.
  11. “Direct Pool-to-Pool Transfers: Curve TriCrypto LP Tokens Have ….” 28 Sep. 2021, https://composablefi.medium.com/direct-pool-to-pool-transfers-curve-tricrypto-lp-tokens-have-been-added-to-our-poc-37598d3afcb8. Accessed 29 Jun. 2022.
  12. “DefiLlama – DeFi Dashboard.” https://defillama.com/. Accessed 29 Jun. 2022.
  13. Vitalik Buterin at Web 3.0 ETH Developers Summit, May 2022. Source: “https://www.marca.com/en/lifestyle/us-news/2022/05/28/6292320e46163ffbb58b4588.html”.
  14. “SingularityNET – Home.” https://singularitynet.io/. Accessed 29 Jun. 2022.
  15. “Cortex Labs: Home.” http://cortexlabs.ai/. Accessed 30 Jun. 2022.
  16. “AI-powered Velas to Launch its Decentralized Multi-Currency Wallet.” 31 Jan. 2020, https://medium.com/velasblockchain/ai-powered-velas-to-launch-its-decentralized-multi-currency-wallet-6d5e4b8dae82. Accessed 29 Jun. 2022.
  17. “Numerai.” https://numer.ai/. Accessed 29 Jun. 2022.
  18. “Sharing Updatable Models (SUM) on Blockchain – GitHub.” https://github.com/microsoft/0xDeCA10B. Accessed 29 Jun. 2022.
  19. “yAI.Finance Rebranding: AI-powered Yield Aggregator Platform.” 21 Jul. 2021, https://blog.orai.io/yai-finance-rebranding-ai-powered-yield-aggregator-platform-a3dce6fce10. Accessed 29 Jun. 2022.
  20. “OptyFi.” https://opty.fi/. Accessed 29 Jun. 2022.
  21. “KingDeFi – Medium.” 2 Jul. 2021, https://medium.com/kingdefi. Accessed 29 Jun. 2022.
  22. “Miljan Martic – Co-Founder & CEO – Kosen Labs | LinkedIn.” https://uk.linkedin.com/in/miljanmartic. Accessed 30 Jun. 2022.
  23. “Peter Toth – Co-Founder & CTO – Kosen Labs | LinkedIn.” https://uk.linkedin.com/in/petervantoth. Accessed 30 Jun. 2022.
  24. “Guven Toprakkiran – Engineering Lead at Kosen Labs | LinkedIn.”  https://www.linkedin.com/in/guventoprakkiran/. Accessed 30 Jun. 2022.
  25. “Gurpreet Singh – Chief Operating Officer – Kosen Labs | LinkedIn.” https://de.linkedin.com/in/gurpreetsingh101. Accessed 30 Jun. 2022.
  26. “Kosen Labs Careers | AngelList Talent.” https://angel.co/company/kosenlabs. Accessed 30 Jun. 2022.
  27. “Kosen Labs Trademark Registration – USPTO .report.” https://uspto.report/TM/97064186. Accessed 30 Jun. 2022.
  28. “Kosen Labs, founded by ex-DeepMind researchers, aims to bring ….” 16 Nov. 2021, https://medium.com/kosenlabs/kosen-labs-founded-by-ex-deepmind-researchers-aims-to-bring-the-benefits-of-advanced-ai-to-the-c8ae04119759. Accessed 30 Jun. 2022.
  29. “Anastasija Verbicka – Operations Manager – Kosen Labs – LinkedIn.” https://uk.linkedin.com/in/anastasija-verbicka-742580127. Accessed 30 Jun. 2022.
  30. “Kosen Labs on Twitter: “The Kosen Labs Agent Program is finally ….” https://twitter.com/kosenlabs/status/1493542005837864963. Accessed 30 Jun. 2022.