Smart solutions leveraging Artificial Intelligence
DiveDeepAI is a full-service consulting firm dedicated to using artificial intelligence and machine learning. We are passionate about helping organizations leverage the power of AI for maximum impact.
With proven track records in the field, our clientele includes for-profit, non-profit, and government agencies. We work with healthcare, retail, telecoms, fintech, and other industries to provide them with valuable insights and tools.
We are a group of Machine Learning experts and Ph.D. researchers who share a passion for developing artificial intelligence (AI) tools and applying deep learning algorithms.
seo.co, Trust Clarity
DiveDeepAI developed Green Residential, a smart application that works on efficient search of property according to the user selected features e.g., number of rooms, days on market, area etc. It uses some existing APIs for the efficient and fast search of the most suitable property. Buyers can look at their desired property by selecting their suitable filters and when they find one, they can get information relating to pricing and taxes.
Moreover, the application supports different filters to facilitate the user e.g., area, division, max price, min price max rooms and baths, min rooms and baths etc.The user can submit an offer and keep track of submitted, accepted and declined offers. Besides this, users can search for the properties related to rent, sold etc. to get an idea of rent and property value in the area. The application assists in automating real estate agents’ work resulting in less time and effort involved in search of a good property.
The smaller the dataset for training, the more it is difficult to have a greater test accuracy. DiveDeepAI worked on an R&D project with a dataset of 100 images with labels for detecting forged parts on identification documents. In simple words, it had to detect the parts where the forgery happened. The team started with using ELA filter to identify different signals for original and forged parts.
The next step was to generate new data from the dataset. This was initiated with creating data from all cropped images, along with the pairs that were both real. This caused the training dataset to be unbalanced as we had more real labelled parts compared to the forged ones. To solve this issue, the team came up with a customized function to create 20,000 images (with added noise and rotations) for training, such that they were 50/50 balanced with the fake/real labels. Later on, the team applied ELA filter to these images and extracted features for these filtered images using RESNET.
Finally, the team trained an SVM (with linear and rtf kernels) for classification. This resulted in an accuracy of 73.9% with a standard deviation of 87.7%. Last but not the least, the dedicated team of DiveDeepAI designed a Flask API to classify forged parts using this limited dataset.
DiveDeepAI developed a platform that generalizes e-commerce products on a website. It has three main features named as Generalize Company Name, Generalize Category and Update Title. Generalize Company Name feature takes an excel or csv file and column name containing company names as input. The similarity of parent company is calculated with other company names and if there is any company with the similar name or part of it then it'll be added as company variant. An additional column name "Updated Company" will be added into the file containing company’s parent name. The result is an updated json file.
Generalize Category feature takes an excel or csv file, category column name and sub category column name as input. It picks the first category name and matches it with the other categories and sub categories. If there is a similarity, that category or sub category will be assigned to the first category name and the same procedure will be replicated.
The result is a json object containing all the categories and their similar and sub categories. Update Title feature takes an excel or csv file, list of words, and column names of mpn, description and brand as input. It will clean the description of each record in the file and will create an 'Updated Title' column in which each record written as "Brand Name - Description - MPN". The output is a json object with an additional column called "Updated Title".
DiveDeepAI developed FORMULATE, an intelligent web application. Formulate is a new document transcription service that uses Optical Character Recognition (OCR) using Tesseract to transcribe data from image scan files and processes said data. It is able to recognize and transcribe 95% or more of all form text.
It supports the various form structures and inherents flexibility allowing easy interface with the existing structures of customers such as databases and excel spreadsheets.
Moreover, Formulate can track usage by user and company with timestamps. This allows future scaling of the service and to predict value provided to the service’s customers. Besides this, Formulate can prevent access from unauthorized users. All attempts to access the software are subject to user identification and password control. If the Organization’s license has run out, the system will prevent the user from accessing the software functionality.
"They offered a unique solution."
DeepDiveAI created a custom interface and integration for a software and financial consulting firm. The goal was to integrate an API into a custom front-end solution with a payment solution.
"They offered a unique solution."
Oct 12, 2021
DeepDiveAI delivered the solution successfully within the outlined budget. During the progress, they communicated regularly and provided updates on the client's tool.
The client submitted this review online.
Please describe your company and your position there.
I am the CEO of an advertising and financial software firm. I oversee all operations.
For what projects/services did your company hire DiveDeepAI, and what were your goals?
We were looking for someone to integrate OpenAI's natural language processing API into a custom front-end solution, complete with payment integration for client-facing and internal use.
How did you select this vendor and what were the deciding factors?
We selected this vendor because of their experience in artificial intelligence.
Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.
They built a custom interface and integration with OpenAI's software platform, allowing our clients to use OpenAI's natural language processing engine for building out and creating content for online use.
How many people from the vendor's team worked with you, and what were their positions?
We had two members of their team.
Can you share any measurable outcomes of the project or general feedback about the deliverables?
The project was completed satisfactorily and with the budget parameters outlined by their team.
Describe their project management style, including communication tools and timeliness.
Regular communication, updates to the build and management via our project management tool were helpful from their team.
What did you find most impressive or unique about this company?
They offered a unique solution. We felt it would be a good additional feature to offer our clients.
Are there any areas for improvement or something they could have done differently?
I think if there were items missing it was because we did not communicate with them as frequently as we should have as we were more hands-off.