Cloud, Drones, IoT, Data Engineering/Analytics, Data Science Solutions

Problem Use Case A leading US university with 10,000+ students, collaborated with Canopus to identify profiles of students with high dropout risk so that appropriate interventions can be planned in a timely manner to maximize retention. Main challenges: Despite spending significant resources & effort on student reach out, the student dropout rate was continuously increasing.

Problem Use Case A blockchain start-up company in US wanted to generate predictions for Crypto currencies trends and market price like Ripple, Bitcoin, Dash etc. for various future time intervals based on past trends, sentiments and statistics. Solution Provided Canopus performed Sentiment Analysis on Twitter data, news data, market feeds in order to identify the

Problem Use Case Canopus collaborated with a start-up real-estate service provider in USA to develop an AI based chatbot to perform and automate the customer services for their real-estate business. Client wanted to implement chatbot to automatically suggest, recommend and show properties to the online visitors. Solution Provided Canopus created an AI based chatbot based

Problem Use Case A well-established tier-2 translation and transcription company in USA wanted to automate their customer document conversion and translation process using machine learning and AI. The objective was to convert PDF into Word document, correct its formatting & style as per requirement and then translate those document from English to Spanish. So far

Problem Use Case Canopus helped a leading Medical Device company that provides chronic disease monitoring solutions. The customer wanted to build an open data ecosystem that would unlock the combined value of patient data, and data locked in various silos, including devices, ERP, SFA, complaint management, and manufacturing systems. Customer was facing following challenges: Real-time,

Problem Use Case One of the Retail and commercial banks, operating an advanced distribution network with 85 branches, 140 ATMs, 32 kiosks, 12.000 POS’s, six web portals, internet banking and a call center; collaborated with to ensure that its payment (debit) card stayed on “Top of Wallet” for more frequent customer usage. Bank wanted to

Problem Use Case Canopus collaborated with a mid-sized company based out of Belgium who works in the drone surveying and performing GIS and Geo analytics on the data captured utilizing the drone flights. Client wanted to perform following analysis on the LiDAR sensor data captured from the drone flight over the forest: Geo-position of Christmas

Problem Use Case One of the leading resource management solution providers who help manufacturers manage their energy and resource costs; collaborated with Canopus to accurately estimate energy resource required per machine and subsequently take corrective action to reduce energy inefficiencies within their production processes. They were facing following challenges: Difficult to use historical machine level

Problem Use Case An Israel based tier-2 fabric manufacturing unit wanted to address the fault management in their plant. They wanted to get an indication about the changes that should be made in the production line to get the higher fabric quality. They also wanted to know by doing some sort of comparison between the

Applications In retail and e-commerce, develop product recommendations based on previous purchases, out of millions of available products recommend only few most relevant ones In telecom, recommend service upgrades based on what is more popular among similar users In healthcare, additional services/products/supplements based on patient’s conditions In automotive manufacturing, recommend exchanges, upgrades and customizations In

The Challenge Perform aspect based sentiment analysis on customer feedback and survey data and identify potential areas of improvement Link this analysis with other data sets like weather data, identify root causes of problems and suggest system improvement changes Canopus Contributions Performed aspect-based sentiment analysis and identified potential areas of improvement Currently, working on linkage

Perform text mining on unstructured data like Social Media, Customer feedbacks etc. to generate actionable insights. What is Different? Perform high accuracy sentiment analysis, classification, topic identification on structured data Use complex algorithms to take care of factors like irony etc. Perform text mining on different languages Perform analysis in real time on Social Media

Applications of aspect extraction and sentiment analysis Service industry: Analyze customer product/service feedback to extract sentiments and insights on aspects – Example: In “The food in the airline was very good”, aspect = food, sentiment = positive, insight = “very good” Transportation: Identify issues faced by travellers Finance: Analyze news feeds to identify sentiments on

The Challenge NDTs are the major cost driver in the call center operations The problem at hand is to apply big data analytics on the data generated through their network and call center operations, and derive actionable insights This leads to a reduction in NDT calls and hence the operational expenses NDT’s Average Handling Time

The Challenge Yield Improvement of Wafer during manufacturing process has a significant ROI potential: Every percentage point is valuable and has high $ impact Wafer film thickness in a specified range determines the yield during the process Producing desired thickness depends on accurately measuring thickness during the CMP process and stopping just in time Canopus

Perform various prediction tasks like forecasting, classification, clustering using state of the art algorithms What’s different? Achieve this by using rigorous algorithms like Logistics Regression, SVM, K-means among others to get higher accuracy in our forecasting Use dimensionality reduction techniques to increase accuracy Forecasting and Prediction has applications in Telecom- predict churn, assign customer as

The Challenge Paid $12.5 billion in Fines. Detecting fraud proactively by monitoring emails, chats for AML, Insider Trading, Violation of Dodd-Frank 600 million messages/month Compliance team of < 30 people to review emails Existing keyword-search systems generated too many false positives Canopus Contributions Delivered a semi-supervised NLP based learning solution Coordinated multiple rounds of reviews

Possible Applications Finance: Detecting anomalous transactions help banks detect and prevent fraud Information networks: Prevents theft of proprietary source code or infection of servers Social media: Detection of compromised accounts, detection of bots generating fake reviews, fake news or fake Twitter posts Healthcare: Detection of a disease outbreak, fraudulent claims and payments, errors in data,

Consists of detecting anomalous behavior of an entity while conducting a transaction or performing some activity “The problem of finding patterns in data that do not conform to expected behavior” What’s different Deliver a robust data science framework to detect anomalous events Ingest and process large amount of data to score anomalous behavior to enable

The Challenge A major Spanish TV network wants to do advertising over TV and radio channels more targeted to specific demographic groups Two different data sources had purchasing habits and media habits These datasets are prepared by different folks for different purposes. Data Quality and Integration to be resolved before any analysis can be done. Canopus

Our Experience For a major media company, we integrated two datasets to correlate people’s demographic characteristics with their purchasing behaviour and habits of watching TV – Improved data quality by custom constraint-satisfaction algorithms for missing value imputation Created the ID layer for a major Telecom Network integrating data from 11 applications and resolving issues like

Possible Applications In Customer 360 applications, data from various applications like Marketo, Omniture, Salesforce etc. is integrated to create a single view of the customer. Data Quality is used to resolve any duplication of the entity names In telecom and many other industries, data completeness, conformity, consistency, accuracy, uniqueness, integrity, and process quality are all

Data needs to be consistent, complete and accurate before analysis can be performed. Otherwise, Garbage In Garbage Out.Data Science is used to de-duplicate data, deal with missing values, make information consistent across the dataset. What’s different Provide an end-to-end data quality enhancement pipeline Data quality score tracking to ensure it is maintained above threshold Data

The Challenge Create a platform to monitor and analyze Customer Feedback Canopus Contributions Feature extraction from unstructured Feedback text. Targeted Sentiment Analysis. Find co-occurring features and feature mentions over time. Identify problem areas based on heuristics combining National Sample Survey data and user demographics. Benefits Realized Client was able to monitor and respond to unstructured

The Challenge Move from position #3 to #2 of world’s largest airlines Improve customers’ perception of both domestic and international flight experiences New CEO’s charter to improve Customer Experience Canopus Contributions Performed aspect-based sentiment analysis and identified potential areas of improvement Linked various data sources and identified root causes of negative feedback. Suggested system level

Built a framework for analysis of customer feedback, e.g., tweets, Facebook posts, customer surveys, etc.Provided semantic analysis of multiple unstructured data sources What’s different integrated several unstructured data sources for further analysis automated sentiment analysis of the customer feedback (positive, negative, neutral, etc.) by analyzing grammatical structure of the feedback, identified aspects of the comments

The Challenge Integrate Data from 10+ different sources like CRM data Identify critical features in the combined data set which impact the propensity for cross-sell and up-sell Create a recommendation model incorporating the data about currently used services as well the features identified in the previous step Run the model and prove improved performance with