“Applied Data Science is about more than mathematical equations and spreadsheets. At Filament we have a wealth of experience in extracting valuable insight from noisy data in order to deploy resilient, production-ready machine learning models to a wide range of problems and industries.”
If you are looking to improve your organisational efficiency by baking your intellectual property into machine learning models, we’d love to help.
Chief Technical Officer
Expert Machine Learning Services
At Filament, we help our clients to make their existing workforces more effective by extracting hidden insights from organisations’ existing data and building bespoke Machine Learning models that work harmoniously alongside humans. From detecting emotions in chat messages to assessing drone imaging for damage in industrial equipment, Filament has a wealth of experience in bottling human expertise into reusable scalable ML models and systems.
Natural Language Processing
Natural Language Processing is the practice of automatically processing and understanding human-readable languages like English, French, Chinese with a machine.
Filament’s NLP team includes ex-IBM Watson, ex-Google NLP and ex-Amazon experts and draws upon state-of-the-art knowledge emerging from academia with expertise and experience using bleeding edge techniques and technologies such as deep neural networks and topic modelling. By leveraging these approaches, we are able to automatically derive meaning from documents and carry out a variety of tasks like detecting emotional content, extracting pertinent keywords and useful information such as names, places and key dates as well as automatically sorting and clustering documents by evaluating their semantic similarity.
Computer Vision, Image and Video Processing
Computer vision describes the family of technologies used to help computers to automatically interpret and derive insight from images and photographs.
The Computer Vision team at Filament have strong credentials in the development and use of image processing and machine learning techniques in order to carry out tasks such as object detection and tracking as well as image sorting and classification. Our Knowledge Transfer Partnership with the University of Essex furnishes the team with world-class support and guidance from leading academics in the computer vision domain. Our team make use of Convolutional/Deep Learning approaches as well as more traditional image processing techniques in order to build models that best fit our clients’ needs.
Structured Data and Regression Analysis
We don’t just work with images and text. Our data science team are able to build models that work directly with numerical data. Such techniques can be applied to a wide variety of use cases. From modelling and predicting forecast financial growth to predicting when an industrial component might fail in dangerous/inaccessible environments.
Some of the most powerful systems that we’ve worked on have been those that combine different types of data together. For example, how does a company share price perform when you take into account stock market transactions and sentiment towards the company on social media?
Big Data and Cluster Computing
In the modern interconnected world, hours of footage are uploaded to video streaming sites every minute and it is estimated that globally 269 billion emails are sent every day. It is therefore not unusual to encounter a data challenge that involves information that is too large to be processed on a single machine within an acceptable time period. At Filament, we have extensive experience in dealing with these so-called “big data” challenges using frameworks such as Apache Hadoop and Spark to perform data extraction and machine learning at scale.
Model Performance and Quality Assurance
It is a well-known fact within the data science and machine learning industry that the quality of the data used to train a machine and the features extracted from that data are just as important as selecting an appropriate algorithm. From neatly organised excel spreadsheets to large volumes of unstructured documents stored on disparate legacy systems, our team have dealt with a huge number of data cleaning and linking scenarios and are on hand to offer expert advise on some of the challenges associated with data collection. We can also advise on some of the best practices for creating new document collections for training models and crowdsourcing.
You can find a free introductory guide to some of the challenges around machine learning and data quality assurance here.