top of page
Artificial Intelligence
Our Specialities
AI is everywhere these days and can be a game-changer, if deployed correctly
ChatGPT has taken the world by storm over the last year or so, propelling AI to the forefront of many people's minds. Now, here at Blackdog we don't believe in hype. We're not going to claim that today's AI is going to completely revolutionize every industry beyond recognition. However, we do firmly believe that it can help boost the productivity of almost any business.
We specialize in designing and implementing tailored AI solutions that drive innovation, streamline operations and maximise efficiency. This may take the form of a bespoke "chat-bot" which can represent the tone and personality of your business, a more efficient way of analysing performance metrics or it may be an AI-powered feature like a horse and rider position analyser.
(Not feeling up to speed on all the terms relating to AI? Checkout our jargon busting guide at the bottom of the page)
Bespoke AI Software
We create bespoke AI solutions, tailored to your business and tech requirements.
Whether it's starting from scratch to train a new AI model or using prompt engineering to integrate existing "off-the-shelf" AI into your app or website, we'll find the right solution for you.
Pricing based on a flat fee structure dependent on scope and complexity.
Consultancy
We offer an AI consultation service.
We can offer an overview of AI itself and advice on how it can be used for business. We're not into hype and we're not into re-inventing the wheel. It might be that the perfect solution to your business's problem already exists as a reasonably-priced monthly service and we can help you get it set up and running.
Hourly pricing structure.
How you can work with us
Our process for creating your AI solution
Initial call
A no-obligation meeting to discuss your project’s requirements and how we are able to bring your vision to life.
​
Delving into details
We'll get into the details of your AI project. There's no one-size fits all solution, so we'll get a good understanding for the needs of your project and how we can tailor the solution.
​
We'll use this information to generate a quote.
​See more info here
AI Training
This stage depends on whether or not we need to create a bespoke, custom AI model for your project. If so, we'll need a curated, labelled dataset for the training process.
​
​See more info here
AI Testing
We'll run a "bare bones" version of the final app to evaluate how well the AI is working. The aim is to test on a wide range of use cases and to determine whether the performance is satisfactory, or whether we need to reassess or retrain the AI.
Design
​
After we've trained the model, we can design the visual identity of the end product. We can do this in-house or work with a designer of your choosing.
Software Integration
Now all that's left is to write the software that will run the AI application. This might be a whole new app, or it could be integrating with an existing website or mobile app.
Testing
​
We'll do a last round of testing, which we will invite you to participate in. The more people that test an app, the better, especially with complex AI.
Release
​
And finally we're ready for sign off and to officially release the AI app!
​
​
Our work in action
30 Plants - live example
An app we're developing with a simple premise; track the number of different plant-based foods you eat each week. We use an instance of Gemini to check that the item a user enters is actually plant-based (bacon does not count) and, if so, which of 6 categories it fits into. By using AI, we're saving ourselves the time and effort it'd take to create an exhaustive list of ALL OF THE PLANT-BASED foods in the world.
You can try out the AI for yourself here and now; simply type a word in the box below, click submit and see what category of plant-based food it falls into (if any).
Some things to note. This is an external AI not explicitly trained for the purpose of categorising plant-based food items, so occasionally it may make mistakes in categorisation. No system is 100% perfect; use at your own discretion (as if you were somehow going to make an important life decision based on the outcome).
To limit misuse (and avoid our racking up massive bills!), we've set a limit on the number of times you can submit an item each day.
Ridesum
This app allows for complex evaluation of equestrian sporting performance with very little hassle, only requiring a smartphone.
We trained and deployed a custom horse and a rider pose estimation ML model using a dataset of images and videos. This ML model finds the location of key joints in each frame of a video, without the need for the horse or rider to wear any special clothing or markers.
Users can choose the live-capture or video analysis option. We were able to embed the ML model on the phone to allow for real time, offline analysis, which is perfect for capturing instant results even in the remote, rural locations where people are likely to be horse riding. We also offer cloud-based analysis on videos that have previously been recorded.
Copyright © Ridesum AB
The models we've created here and the analysis used are for equestrian sports, but the same premise can work for many of different sports and applications.
Enduro Fit
The Enduro FIT app works in tandem with their patented equine heart rate monitoring girth wrap to give riders true fitness insights.
We trained and deployed a ML model to detect the gait of the horse based on the movement of the rider's phone. This was a challenge as we know that in real-world use the phone could be placed in a number of different locations (including upside down inside a boot!) plus different types of horses move very differently from one another.
The gait is used in several types of analysis and is shown colour-coded on the ride tracking location map.
Copyright © Ridesum AB
Image copyright © Siametric Systems Ltd
As for the the pose estimation example above, the models we created for this app are equestrian-focused, but the same premise can be applied to lots of other sports and applications.
Tailoring our solution to your business
We have real-world experience of using both Generative AI and Machine Learning in applications. Our aim is to build the right system for your business. Some of our considerations we will discuss with you during the design phase:​
​
​​
-
Off-the shelf or custom model
For some applications, we can short-circuit the development time by making use of third party solutions, like ChatGPT or Gemini and using prompt engineering. This works for more general knowledge tasks, such as how we've used it in our "30 Plants" app to determine which plant category (if any) an item entered by the user falls into. We've also used it for complex character (printed and handwriting) recognition. Other more niche applications, such as identifying the gait of a horse from the movement of the rider's phone or creating a chatbot with expert knowledge of your business will require the training of custom machine learning (ML) models using a carefully curated data set. We have experience of training these ML models from scratch to deployment and the secret lies in getting the right data set.
​
​
-
Cloud or on-device deployment
internet connectivity or privacy concerns will determine if the AI or ML model will run remotely in the cloud or if it ships with the app. Cloud deployment can be more efficient in terms of the size of the app and can add an extra layer of security if using a proprietary model. However, it requires an internet connection (not always a given even today!) and the use of external servers or third part services.
​
​
-
Traditional machine learning algorithms or neural networks
let's try and keep things as simple as possible. While models that use neural networks are extremely powerful, they're also computationally expensive and need large data sets. So we'll try and use the most efficient type of AI that we can to get the optimum result. There's no point in draining a user's phone battery with an over-engineered behemoth of a model that took us a month of compute time to train when we don't need to!
​
​
-
Existing or new app/website
Do you already have an app or website that you're looking to integrate a new AI feature into? If you're looking for a completely new system, we offer both mobile app and web development. It's also an option for us to just develop the AI or ML model and turn it over to your developers for implementation in existing code.​​​​
​​
​
-
Access to training data
If we need to create a bespoke, custom AI model from scratch, we will need a labelled data set for training. Do you have a data set or is this something we need to create and label as part of the build?
​
​
​
Don't worry if this feels a bit overwhelming - we will walk you through everything and are happy to explain anything you're unsure of. Please never be embarrassed to ask questions, there's a lot of jargon in the AI field. Hopefully our guide below will help you decipher a few things in the meantime!
anchor details
LLM
Large Language Model
This type of AI has been trained on vast amounts of data (e.g. the entire internet) and is designed to take input in the form of normal human language, rather than code. It also generates human-like responses. ChatGPT is an LLM.
AGI
Artificial General Intelligence
This is basically referring to robot intelligence. It's AI that's capable of understanding, learning, and performing any intellectual task that a human might be able to (and more besides). This is the goal many groups are working towards, but no-one has achieved it so far (or as far as we know!)
Handy jargon busting guide
NLP
Natural Language Processing
This is the bit of AI development that focuses on enabling computers to understand, interpret, and generate human language as opposed to software code.
ML Model
These models are trained on lots of data (like images) to identify patterns (e.g. what a duck looks like) and make predictions or decisions (e.g. is it a duck?) without being explicitly programmed (e.g. a programmer has not described a duck using code, just lots of images of ducks).
Neural Network
This is a type of computational model inspired by the structure and function of the human brain, consisting of interconnected layers of nodes (i.e. neurons) that process data and learn patterns to make predictions or decisions. They're a bit like a black-box; you can't open them up and find a line of code that does a specific task like you can with traditional software.
Algorithm
An algorithm is simply a step-by-step set of instructions or rules designed to perform a specific task or solve a particular problem. In software, these steps are usually blocks of code.
Predictive AI
Also referred to as Traditional AI or Machine Learning (ML), these systems are great at specific tasks. For instance language translation, object recognition or pose detection (i.e. detecting a person or animal's posture)
These systems require specific training on carefully curated datasets created for a single purpose or task.
Generative AI
Gen-AI refers to systems like ChatGPT, Gemini or Midjourney which creates new content such as images, text, music or videos based on an input prompt provided by a user. Large language models (LLMs) are Gen-AI.
These systems are trained on vast amounts of data (i.e. the ENTIRE internet plus media and artworks), which has given rise to arguments about copyright and how "new" the content these system produce really is.
Deployment Platform
This basically means where will the AI software run. Broadly speaking there are two options; on device (i.e. directly on the phone) or in the cloud.
There are many factors to take into account when deciding whether on-device or cloud deployment is the best solution, including model size and complexity, internet access, privacy and cost.
anchor jargon
Next Steps
The information and examples on this page are only a very small glimpse of what we're capable of building with AI and code.
If you like what you've seen and read, it's time to get in touch with us. Feel free to send an email, WhatsApp or call with any questions you might have or to set up an initial meeting.
​
We love a challenge and to push the limits of what's been done before, so absolutely hit us up with your crazy ideas!
​
​
bottom of page