Barclays India Women’s Hackathon 2019

The problem statement it to find the negative sentiment about Barclays on Social Media Platforms such as Facebook, Twitter, etc.

Business Outcome:

• Identifying and targeting the dissatisfied customers to ensure the non-proliferation of the negative sentiment and avoid customer attrition.
• Identifying the dissatisfaction related to a product or a specific feature

Technologies required:

• Python/PySpark
• Knowledge about NLP and ML

Which Account should I use to consume social media feeds? Use your own Developer account to create webhooks for twitter, facebook etc to listen to the feeds.

Which library to use? Use any 3rd party Library or Cloud services for sentiment analysis and classification.

Technology to be used – Machine Learning / Data Science - Python/Java/SCALA

The problem statement is to build a predictive model for attrition which will help the organization in better resource planning and improve the employee experience.

The data elements that can be considered for building the predictive model can include the following:

• Performance ratings – Outstanding / Strong / Needs improvement
• Educational background
• Time in title
• Company tenure
• Industry experience
• Role
  
• Area of expertise/skills
• Career progression
• Compensation placement – alphanumeric
• Market demand (will need to be sourced from external agencies)
• Talent pool (will need to be sourced from external sources)

The problem statement is to build a feature/code/tool/portal, where UNIX servers can be registered for reboot and it should record following points:-

• Basic initial checks (cpu, memory, file system size, network links, network routes, vas status, some basic/ecosystem processes)
• Reboot Status
• Post reboot checks in similar format
• Should include kdump/console logs
• Read the schedule time and register the server

This can be achieved on any cloud platform, as this feature is not available on AWS natively. The report should compare pre and post boot status and report in below format.

• Web Screen should have options to upload 2 server list files with similar data attributes as above. (PS: above columns are sample, real columns will be similar to above)
• There should be an Option to call the Web service to check the status of the server or servers.
• Preferred Technologies usage: Python for operation logic and Angular JS for Web application Working code output required.
• Solution is usable by business: Yes

Additional Requirement Tips:

• How do we display the stats of an individual server? Display the stats in tabular and graphical format on click of individual server in the above table.
• What commands do I use to get the metrics stats? Use Unix Commands or you can consider using Cloudwatch commands if using AWS cloud
• Where do I store the server list and its stats collected? Store the data in AWS Cloud s3/NoSQL/RDS
• What are data archival requirements? Data Older than 6 months should be auto-archived and should be made available within 24 hrs if requested.
• Can I set up and configure the environment manually? You can start the solution development manually, however, It is good to have automation including CI/CD and Infrastructure as a Code through Terraform/Chef or Cloudformation.
• Who will bear the cost of the services used in Cloud? Only services available in Free-tier to be used and resources needs be terminated as and when you are done or by end of the day. DO NOT leave your resources running. Barclays will not reimburse any cost incurred and recommend you to use only free-tier services.
• Architecture of the solution design: Your web application should be highly available, scalable, cost effective, High performance and Fault tolerant. Use Cloud server-less managed services wherever applicable

1. Restful Web API to be written for Probability-based algorithm to be applied for potential data-attribute value matches as a match criterion. e.g. where these sample Banking trade data attributes (columns) :
   • trade_value_date,
   • legal_entity
   • Trade Volume
   • Currency_name
   • transaction_value Should match potentially across 2 different datasets.
2. Above fuzzy matching logic linking “text” to “similar text” should also be applied to combine datasets with uncooperative columns (i.e. columns that are not obvious but have same underlying meaning).
   
   
3. Web Screen should have options to upload 2 data files with similar data attributes as above. (PS: above columns are a sample, real columns will be similar to above)
4. There should be an option to call the Web service to prepare the matching criteria that will then suggest matches on screen.
5. Prepare a self-learning Dictionary where User can be allowed to add examples like below to Dictionary:
   • Manual Growth of Dictionary: “International Business machines” should be used in future as “IBM” also after User accepts first time.
   • Automatic growth of Dictionary: as per sample outputs above.
   • Dictionary should maintain growth in database for persistence, not in memory.
Technologies to be used: Python or Java for machine learning and Angular JS for Web application Working code output required. Solution is usable by business: Yes