Humans create, machines work.

We are a group of scientists, engineers, and developers who are passionate to revolutionize the future of businesses with AI and machine learning technologies. 

We provide custom AI solutions and products to help companies across different industries to implement AI today. Our AI research and development facility allows us to solve today’s complex problems, customize real-life AI applications for positive future impact, and drive future innovation.

 

We are on a mission to develop AI products and solutions that unlock human potential.

 
$5,000+
 
$50 - $99 / hr
 
10 - 49
 Founded
2019
Show all +
London, United Kingdom
headquarters
  • International House, 64 Nile Street, London, N1 7SR,
    London, LND N1 7SR
    United Kingdom

Portfolio

SemanticForce: Aspect-Based Sentiment Analysis Image

SemanticForce: Aspect-Based Sentiment Analysis

The Challenge

Analysing millions of public comments on social media is important for marketing campaigns to help companies to understand better what exactly clients like or dislike in the products. Currently, this work is done or by hired experts who have some elementary domain and language understanding or is outsourced to third parties. The first option is getting expensive on the scale since it

requires more and more people that also have to be trained. The second option is risky since such gathered data has R&D value and can't be shared very easily. Using natural language processing (NLP) algorithms leveraged by the latest development in AI, especially deep learning and state-of-the-art results in natural language understanding and generation (NLU, NLG) seems like a great fit for the given mission.

The Solution

Typically, the classification of a piece of text into different emotional or sentiment groups is called sentiment analysis in the literature and it is well solved with the latest AI developments. Although, the classical sentiment analysis doesn’t fit here since it just tells the average tonality of sentences instead of a detailed explanation of people’s preferences about entities and words that appear in the text. For instance, how would you classify such a sentence "Thanks, the coffee was good, but I had to wait for too long for it"? Positive, Negative, maybe Neutral class? In fact, you would rather say that the coffee thing is positive, but service is negative. This slight change of concept arises into a totally new task class called aspect-based sentiment analysis (ALSA).

The solution we have built was a completely novel NLP system for fine-grained aspect-based sentiment analysis, topic modelling, and text categorization for different languages and data sources including automated machine learning pipeline. The core of it was the beforementioned BERT deep learning model that was trained to detect different entities, their sentiment polarity, and corresponding trigger words.

The Result

The final value of the project was in cutting costs and optimising time and process for social media analysis with optimising routine work for more than 30% of the staff with a fully automated AI pipeline.

We have observed not just the replacement of the manual work, but also an overwhelming speedup of the work: a dataset that could have been processed for weeks by the human team is processed by the algorithms in several minutes.

Image Logo Retrieval Image

Image Logo Retrieval

The Challenge

Brands are being highly aware of their appearance in social media, but not everyone actually uses hashtags or even mentions the brand itself in the post. Meanwhile, companies would like to know what the audience is talking about them and in which context. Are they satisfied with the product? Who are they sharing their emotions with? Where does it happen? Posts with images or videos are

perfect for analysing from this point of view, but it takes too many resources at the constantly growing web-scale. As an analogy with the fine-grained sentiment analysis case described above, the main challenge was in developing such a solution that can replace human staff. This solution has to save resources like time, money on the scale.

The Solution

Computer vision technologies are mature enough to solve the given task: algorithms like YOLO or SSD given enough data can show needed accuracy. The challenging part was in adaptation for different brands, markets, and image conditions: we would save time and money also on the data gathering and labelling processes. This is the point where we had to dive deep into R&D and adopt state-of-the-art solutions.

We have developed a deep learning-based image retrieval system that can find any given logo of any brand in millions of images and videos from social media. The average accuracy was higher than 85% and the overall system was scaled up to process 10k+ pictures in just a couple of minutes.

The Result

The analytics department has reduced its load, and costs at 35% and B2B clients of our customer have reported better quality and wider coverage of the situations in the marketing reports. We have observed not just the replacement of the manual work, but also an overwhelming speedup of the work: a dataset that could have been processed for weeks by the human team is processed by the algorithms in several minutes!

Ilogos: Player Analytics Model Image

Ilogos: Player Analytics Model

The Challenge

Today’s videogames market is full of competence, and it’s not a trivial task to keep players engaged in a particular game and, moreover, spend money on the playing benefits. Since there are a lot of games that are very similar to each other by the gameplay and customer journey, it is crucial to fight in any possible way for your customer and its LTV. Very often, it involves sophisticated

marketing campaigns and inside-games funnels that are carefully created by game designers. One of the other ways to keep players interested is to adapt a game dynamically to their mood and state. But is it really possible? Can AI help with it?

The Solution

If we are talking about analyzing the players' psychological state, you might think or of psychological tests and polls to estimate a person's engagement while playing, or even measuring biosignals like pulse and brainwaves to get immediate feedback about what a player feels in each exact moment. Both of these approaches will take too much time from the biomedical experts. In the first case, the feedback won’t be immediate. In the second one, we will meet additional expenses for the measurement devices.

We provided a complex solution to our client who asked us to solve this problem without any psychological data available at the training stage, just for the validation of the approach.

The technical solution itself was a novel mathematical model on the intersection of data science and cognitive modelling that became part of the intellectual property of our partner. Based on the high-level player's actions in the game like general activity and purchases and very low-level details such as information from the phones accelerometer, camera, and microphone we could recover the patterns that are correlated with bio-signals. The model provides a game designer with a single number, which is a predictor of churn, in-game purchases, overall performance and satisfaction with the gaming process and serves as a “playing pulse” of each of the millions of players of the game in real-time.

The Result

We helped our client to enter the market of adaptable gaming and reduced players’ churn by 10%. Our solution increased the average time spent in the game with 15% and inside game purchases by 3%.

Also, it allowed a game designer to understand people's behaviour much better in real-time which allows developing a totally new game with new mechanics and monetization.

   

 

 

MLVCH: style transfer for photos and videos Image

MLVCH: style transfer for photos and videos

The Challenge

People are always trying to look cool and original in the feed, that’s why filters that make the photo better or masks with makeup or animal faces are booming last years. This trend doesn’t seem to be stopped, only being deeply adopted by Facebook, Instagram, Snapchat and other big companies. For our client, we have developed an app that can repaint your photo or even the video in the style

of any famous artist in the smallest details!

The Solution

We have deeply researched and implemented several approaches to the image and video style transfer and adapted the algorithms for the visual and artistic feel of our client. The open-source available solutions have several bottlenecks and pitfalls that make it not so straightforward to apply for real-world use cases. For instance, the granularity and sizes of the "strokes" drawn by the AI-artist require some tricks to be controlled. Another issue is in video generation - it's easy to process it frame-by-frame. Still, in the end, the video has to be smooth, which made us solve an additional optimization problem. We adapted and improved known solutions according to the product point of view and the final solution was based on deep neural networks and could run on mobile devices in real-time.

The Result

Our client had launched a competitive product being one of the first on the market that had 100K+ downloads on the mobile app markets and adapted the solution for successful B2B sales for marketing purposes.

Apart from the working product itself, the significant applied research knowledge base was created related to the artistic style transfer technology development, which allows building new related products very quickly and avoid possible pitfalls.

   

 

 

Automated trading systems Image

Automated trading systems

The Challenge

Our customer is a trading boutique that wanted to improve its performance in the cryptocurrency market. In purpose to minimize non-systematic trading, the customer was looking for a quants team with experience in AI-based algorithmic trading. The main goals were:

  • Yearly Performance > 25%;
  • Maximum Drawdown < 15%;
  • The system had to be fully
automated;
  • Diversification (system needed to consists of different uncorrelated strategies)
  • Hedge (system needed to perform on both bullish and bearish markets)
  • The Solution

    With a team of 2 quantitative researchers, 2 quantitative developers, 2 traders, 1 DevOps, 1 quantitative risk manager and 1 product manager we combined our AI and trading experiences, researched publicly available research papers and developed and deployed next strategies:

    • Alternative Data Strategies;
    • Trend Following Strategies based on Machine Learning algorithms;
    • Statistical Arbitrage Strategies;

    To train models, our team developed infrastructure for collecting both market data from top exchanges and alternative data from news websites, Twitter, Reddit and other social networks.

    Because of unstable crypto market infrastructure and the existence of technical issues on a lot of crypto exchanges our team developed 'Market sandbox’ for bots testing to ensure that behaviour of the bots in real trading would be the same as on backtest.

    The system was designed in a way which allows us and customer easily add new strategies and trading bots. Also, it was designed to be fully automated which means that it buys and sell cryptocurrencies by itself.

    For scaling, we connected the system to more than 10 top crypto exchanges (Binance, Bitfinex, OKEx, BitMex, Poloniex, Huobi etc.).

    For secure and effective asset management we developed for the customer:

    • Risk management system
    • Monitoring and notification system

    Crypto markets work 24/7, so to manage and maintain hundreds of trading bots on different exchanges our team of engineers and DevOps developed control, risk management, and notification systems.

    The Result

    The customer got a multi-strategy system with several trading strategies, which are uncorrelated, because of extreme different logic of decision making. The end solution is quite easy to scale, maintain, modify and monitor. The whole system performance can be evaluated in the next metrics:

    • Average monthly performance - 1.96 %;
    • Maximum Drawdown - 6.2 %;
    • Sharpe Ratio - 2.1
    • Sortino Ratio - 3.4

    Reviews

    Sort by

    AI Tool Development for Mobile Game Studio

    "Even though Neurons Lab’s expertise is in technology, they understand the project from a business perspective very well."

    Quality: 
    5.0
    Schedule: 
    5.0
    Cost: 
    5.0
    Willing to refer: 
    5.0
    The Project
     
    $10,000 to $49,999
     
    Feb. – Apr. 2019
    Project summary: 

    Neurons Lab was hired by a mobile game development company to build an AI tool that tracks data related to their game players. 

    The Reviewer
     
    201-500 Employees
     
    Hamburg, Germany
    Nikolay Minaiev
    CEO, iLogos Game Studios
     
    Verified
    The Review
    Feedback summary: 

    The final development is on hold for now, but so far Neurons Lab’s tool has performed perfectly during internal testing stages. Customers can expect a well-rounded team with exceptional communication skills.

    A Clutch analyst personally interviewed this client over the phone. Below is an edited transcript.

    BACKGROUND

    Introduce your business and what you do there.

    I’m the CEO and head of production of iLogos Game Studios, a mobile game development service provider. 

    We offer full development, art production, live operations, and porting for our partners and customers. However, our main focus is free mobile games.

    OPPORTUNITY / CHALLENGE

    What challenge were you trying to address with Neurons Lab?

    Our internal R&D project required the development of an AI-based tool that could track a game player’s behavior and predict their actions.

    SOLUTION

    What was the scope of their involvement?

    We described our idea and the goals we wanted to achieve, and Neurons Lab collaborated with us to build the solution from the ground up. We provided our own business perspective of the concept, and they were able to create the tool using mathematical models and other data capturing strategies. 

    What is the team composition?

    We worked directly with their CEO, who was our main point of contact, and one senior data scientist. 

    How did you come to work with Neurons Lab?

    They were recommended by our partners.

    How much have you invested with them?

    We spent $20,000 on this engagement.

    What is the status of this engagement?

    The project took place between February – April 2019.

    RESULTS & FEEDBACK

    What evidence can you share that demonstrates the impact of the engagement?

    Though we’re currently postponing the development of the tool itself, the product we’ve received so far works perfectly. We’ve been able to compare real data with our predictions in internal testing, and we’ve received great results. 

    How did Neurons Lab perform from a project management standpoint?

    Neurons Lab’s project management is very effective. We had daily calls to discuss progress and planning, kept in touch with Skype and email, and tracked reports in tasks in Jira. We worked together almost in real-time, and never had any delays as a result of communication.

    What did you find most impressive about them?

    Even though Neurons Lab’s expertise is in technology, they understand the project from a business perspective very well. They have an excellent ability to connect tech with our final goals as a company.

    Are there any areas they could improve?

    No, everything was perfect.

    Do you have any advice for potential customers?

    Preparation is really important, so spend the proper amount of time in pre-production. Think about your goals and the tools you’ll use.

    5.0
    Overall Score
    • 5.0 Scheduling
      ON TIME / DEADLINES
      There was full transparency on the production and report sides, and there were no delays with communication.
    • 5.0 Cost
      Value / within estimates
      AI-based work is really expensive but they were worth the money.
    • 5.0 Quality
      Service & deliverables
      Everything was on time and as planned. It all worked and was delivered with the proper documentation.
    • 5.0 NPS
      Willing to refer

    AI Development for Cryptocurrency Company

    "They met the goals we needed."

    Quality: 
    5.0
    Schedule: 
    5.0
    Cost: 
    5.0
    Willing to refer: 
    5.0
    The Project
     
    $200,000 to $999,999
     
    Dec. 2017 - Dec. 2018
    Project summary: 

    Neurons Lab provided AI development services, working on various platforms to improve trading systems. They used Python, C++, and machine learning tools for the development work.

    The Reviewer
     
    11-50 Employees
     
    Cayman Islands, United Kingdom
    Co-Owner, Trading Boutique
     
    Verified
    The Review
    Feedback summary: 

    The team delivered consistent results that met expectations at all times, particularly with regard to the investment metrics. They're good collaborators who hold regular meetings and reviews to ensure excellent communication throughout.

    A Clutch analyst personally interviewed this client over the phone. Below is an edited transcript.

    BACKGROUND

    Introduce your business and what you do there.

    I'm one of the investors and stakeholders at the quantitative trading boutique.  We trade on the cryptocurrency market using automated trading systems.

    OPPORTUNITY / CHALLENGE

    What challenge were you trying to address with Neurons Lab?

    We hired Neurons Lab to develop the automated trading system with trading algorithms, risk management and monitoring system. This ATS had to generate for us at least 25% per year.

    SOLUTION

    What was the scope of their involvement?

    We asked Neurons Lab to develop several trading algorithms for us for better diversification and the ability to generate returns on both bullish and bearish phases of cryptocurrency markets. The system needed to be very stable because we were planning to attract big investments for our fund. Neurons Lab developed for us a few algorithms: news-based, sentiment analysis (using NLP), based on alternative data, trend following based on machine learning, and arbitrage & statistical arbitrage. For secure and effective money management they developed risk management, monitoring and notification system, and order execution system for more than 10 top exchanges.

    What is the team composition?

    I think there were seven or eight people: the CEO, CTO, two quantitative researchers, two quantitative developers, and one or two business analysts.

    How did you come to work with Neurons Lab?

    We spoke to different companies and teams that could help us with development. Neurons Lab has some performance on the US stock market, and it was necessary for us to find someone experienced in the cryptocurrency market.

    How much have you invested in them?

    We spent around $300,000.

    What is the status of this engagement?

    We worked together from December 2017 till December 2018.  Time-to-time we ask guys from Neurons Lab to help us with risk management and model evaluation.

    RESULTS & FEEDBACK

    What evidence can you share that demonstrates the impact of the engagement?

    It is all about the investment metrics. The average monthly performance is 1.96 %. The maximum drawdown was 6.2 %. The Sharpe ratio is 2.1.

    How did Neurons Lab perform from a project management standpoint?

    In general, everything was good. We had weekly meetings and planning sessions, as well as reviews with one meeting a week. They did a good job of understanding our requirements and we're satisfied.

    What did you find most impressive about them?

    They met the goals we needed.

    Are there any areas they could improve?

    Everything was good.

    Do you have any advice for potential customers?

    They're good and they know what they're doing. Anyone working with them will have a good experience.

    5.0
    Overall Score
    • 5.0 Scheduling
      ON TIME / DEADLINES
    • 5.0 Cost
      Value / within estimates
    • 5.0 Quality
      Service & deliverables
    • 5.0 NPS
      Willing to refer