Use cases
In France, every year, 150,000 contracts with a unit value of more than €90,000 excluding VAT are awarded in public market tenders for an average amount of €90 billion. This concerns many areas of economic activity. Many companies spend a lot of time studying and analyzing them to decide whether or not to respond, with no guarantee of results. Their statistics, if well maintained, give them a first indicator.
With Machine Learning, we go further. It allows the integration of additional data to correlate the elements requested in the market tender with the characteristics of the company, while taking into consideration the evolution of these characteristics and also of external data.
Improved process and productivity through automatic analysis of market tenders.
Access to the scoring and decision support tool for all employees for a collegial analysis and minimization of errors of appreciation.
Determination of a prediction score for winning tenders.
Necessary data history on past tenders and their results
Determination of the completeness of the data concerned in the RFPs and the respondent company’s data
Update over time the evolution of the characteristics (strengths and weaknesses) of the respondent company to avoid model biases
Build and train a machine learning model
Real-time automated prediction generating a confidence score of the ability to win the market tender
Study time divided by 10 – Potential business flow multiplied by 5
Use cases
Eating disorders affect 1 million people (mostly women) in France, including 600,000 young people under the age of 35. 2 to 5% of the world’s population is affected by its various forms: overeating, bulimia and anorexia. These disorders lead to health complications throughout life and increase the risk of early death. Diagnosis is hard due to the difficulty that patients have to talk about it.
The artificial intelligence implemented thanks to AI Parc will allow to automate the detection, the prediction of risks, the assistance in choosing the care journey and the patient support.
Detection, diagnostic support, assisted analysis
Patient pathway analysis
Accompaniment of the patient in good eating behavior
Difficulty in making a diagnosis and defining a treatment
Recovery of data
NLP semantic analysis for Bert model creation of a patient self-evaluation Bot
Use of neural networks
Patient journey analysis by statistics and Machine Learning
Use cases
In the context of improving the performance of a community’s apprenticeship training policy, the analysis gaps between the needs and the apprenticeship contract offers is significant.
The Machine Learning implementation to analyze the gaps between the supply of contracts and the demand for apprentices is based on 4 criteria:
the profession
the level of diploma being acquired
the geographical area
the age of the apprentice
The visualization application helps to quantify and qualify the gaps between supply and demand in order to decide on appropriate measures to reduce these gaps.
Creation of a predictive tool to help choosing an apprenticeship policy
Visualization of offer and demand curves of apprenticeship contracts with applicable criteria and filters
Self-assessment of the forecast quality (allowing to detect possible missing data)
Progressive integration of new data sources
Time to analyze and extract useful data from the learning contract databases
Data analysis via EleasticSearch and development of a multi-criteria analysis prototype.
Use of Machine Learning, time series, clustering and forecasting using Prophet
Model creation in a collaborative IA PARC project, published, validated by the test team and scaled for deployment on a curve and prediction data visualization application
Use cases
In a manufacturing unit producing parts for the automotive industry, 3×8 operators visually inspect parts at the end of the manufacturing line. Their task is to remove machined parts with defects.
The work is tiring, not rewarding and performance is limited. Detecting failures is also difficult.
With Deep Learning, it is possible to detect and remove automatically defective parts at the end of the line.
Valuation of the work of the operators at the end of the production line by making their task evolve
Automation of the control by fixed cameras connected to a control PC carrying out the analysis in real time of the images with a pre-trained AI.
Improvement of detection performances
Lack of data history
Generate large amounts of data using gantries equipped with cameras to take pictures of 10 features of defective parts
Automatic learning on the IA PARC of several hundred shots from different angles and by type of defect possible
Real-time automatic detection of defective parts coming off the production line
On-site installation of a dedicated IA PARC
Processing software with operator station and camera control
Use cases
The project targets industries wishing to implement supervision and decision support tools related to predictive maintenance.
Fastpoint and MomentTech are combining their expertise to collect and deliver data from industrial equipment (example : radiation protection measurement equipment) from the most constrained industrial sites in order to analyze and exploit it on neural network models that provide better results.
The NucleoT solution objective is to inform operators and subcontractors on the predictive maintenance actions to prioritize.
Ability to predict equipment failures and malfunctions as early as possible
Gathering, securing and transmitting data from radiation protection equipment
Weak signals identification in radiation protection
Reliability of data interpretation made possible by deep learning algorithms
Implementation of synchronization mechanisms for models/data sets used by AI
Build of scalable and customizable predictive models on radiation protection data
Models training and optimization
Use cases
Facial recognition: an obvious issue in the field of security.
A security company wants to allow its camera network to continuously film the crowd in a shopping mall in order to identify a wanted person.
Image recognition is less known in the health field, however it represents a great need, especially in the detection of diseases or advanced diagnosis through radiological images or analysis results.
Simultaneously and in real time, analysis of all faces in a space
Filtering by criteria and extraction of approximate faces
Increase of the resolution to improve the sharpness of the face
Comparison with a database and identification
Multiplication of cameras to be installed as the crowd moves in all directions
Recognition of an individual hiding behind an attribute
Model training
Use cases
Contingency planning is implemented by public authorities, emergency services or private companies. It consists of creating a mechanism of anticipation to respond to any kind of crisis: a war, a pandemic, a natural or technological disaster (factory fire, cyber-attack, faulty or dangerous product…). Contingency planning helps mobilize the necessary resources to ensure the safety of people and property.
AI will enrich the historical data of conventional planning with more advanced Machine Learning algorithms while taking into account internal data, and external data for a detailed analysis of variables.
The Covid pandemic was a major field of application for machine learning, particularly for predicting the spread of the virus, communicating and providing relevant information remotely, developing telemedicine and ensuring food supplies.
Detection of Covid and prediction of its expansion using Machine Learning algorithms
Anticipation of the consequences of its propagation on the whole social, economic, health and institutional activity of the country
Collection of data contained in the health reports of 65 countries and as many languages, the location of animal diseases, added to the data of airlines.
R tool implementation which, in addition to the language, provides an environment for data processing, modeling and visualization.
Manipulate data by integrating a series of statistical, clustering, classification, analysis and graphical techniques
Finding and optimizing the best model for an easy and intuitive implementation made possible by choosing the best Machine Learning frameworks: Scikitlearn, Tensorflow or Pytorch