PATHWAY
SAMPLE PS
2018
OCTOBER 20
1
Introduction
To help us better understand your skills and approach to the problem statement, we
have prepared the following list of tasks as part of your mock assignment.
We encourage you to approach it carefully and use it as an opportunity to showcase
your technical aptitude and creativity
Please note that task 1 is compulsory while at-least one of task 2-4 may be
attempted.
All the best
2
Task 1:
Setting up Docker
Pathway's library only works on Linux environments (unfortunately for us poor
Windows users). Hence, we need to set up docker containers in order to ensure that
every person on the team works on the same environment. So, your first task is to
1) Download Docker Desktop ([Link]
2) Learn how to deploy Docker for a pathway application
([Link]
3) Show the execution of a program using the pathway library after applying docker
containerization. (In this case any sample program will be accepted)
Now that you know how to deploy docker you can start working on the next part
3
Task 2:
Real time AI-driven stock price prediction
Over the last few years multiple companies have been leveraging AI in order to
improve the quality of investment decisions. Since a fully-fledged AI based stock
market predictor is beyond the scope of this mock problem statement, you have to
implement a toy example in which a program can make financial decisions with
minimal human intervention. To do this you have to complete the following
1) Train an AI model on historic data.
2) Apply the model on a market simulation where the model uses the price of the last
few timestamps in order to predict the future change in the stock value.
3) Show a plot of predicted price vs actual price along with an accuracy score. The
accuracy score is defined as the percentage of times the model predicted the
direction of the stock (Up or down) correctly.
Note: - The model CANNOT look into the future price while making a prediction. (The
prediction for timestamp tn cannot take tm for an input if m>n). In case your program
does look at future data, it will IMMEDIATELY be disqualified. We place no
restrictions on what data you use to in order to make your predictions as long as it is
related to finance.
Sample Data Link: - [Link]
4
Task 3:
Smart KYC checker
In this task you have to implement a barebones AI-powered KYC checker. Your
solution should be able to do the following
1) Parse user documents (Like Aadhar, PAN, etc.).
2) Extract relevant user information from these documents.
3) Perform basic fraud detection (E.g. Checking names and details across documents).
4) Bonus points for implementing OCR (Optical Character Recognition) instead of
reading text documents, since most people upload photographs.
Additional features will fetch bonus points for creativity.
Task 4:
AI-powered Customer Support
Automating customer support is one of the key areas of application for artificial
intelligence. In this task you have to build an AI system which will communicate with
a user in order to answer any financial queries it may have. The requirements are:-
1) Take user input in the form of a string.
2) Process the user input in a language model (may be hosted locally or on the cloud)
and generate a response.
3) Display the user response.
4) Bonus points will be given for storing important user details from their text
prompts.
5
Submission Guidelines:
Your submission must consist of the following things
1) A GitHub link containing with your code.
2) A 2-page report with additional information about your approach.
As previously stated, you don’t need to do all the tasks in the problem statement.
We will prefer quality over quantity.
Please note that you will be asked to explain your code during the interview.
Submission Link: - [Link]
Submission Deadline: - 11:59 pm 22nd October