Karthikeyan Kadirvel

Data Scientist with 7 year experience in SCM domain

Chennai, India

Summary

I am having good exposure to Supply chain management domain and in finance (Algo Trading)

Languages:

English, Tamil

Favorite Python Packages:

Pandas, Numpy, dask, Networkx, Dash, sklearn, tensorflow,keras

Experience

Project #1: Sequence mining in the click stream data

Objectives: To understand the user behavior using click event data and to predict the order probability and next action of the user.

Problem faced: Couldn’t able to identify the bottleneck click events and couldn’t track of market campaigns sales exactly. To give better experience to the visitors

Solution: Using NetworkX build the EDA analysis and using gephi software given visualization of the event click data. Using LSTM- deep learning created the model to predict the order probability and next action of the user.

Description:

• Using python connected to Teradata and get the data
• Embedded the click events in the network graph and developed different metrics
• LSTM model deployed using tensorflow and keras

Project #2: Developed a feature engineering product

Objectives: To develop the feature engineering project, that will help the data scientist to get the data and to do the initial EDA analysis

Problem faced: Since many data scientist working on various projects to build the model, there is always duplicate of huge data. And initially to get the data and do the basic cleaning work items data scientist spending more time on it rather than of hyper tuning the model.

Solution: Created a user interactive web application where user enter the database details and SQL quires. So that this tool record of it help the user. And also we provided the a API link so that user can get the data directly.

Description:

• Using flask created the web application of this tool
• Pyspark to process the data.
• NoSQL to store all the data.

Project #3: Root Cause identification of OTD fail items- Decision tree algorithm (ESAB)

Objectives: To improve the OTD of spares items

Problem faced: OTD metric are at the low level

Solution: To proactively work on the sales order to avoid the OTD hit. Developed the forecast model using decision tree algorithm whether to work on the SO (sales Order) proactively or not.

Description:

• Using R connected to SQL for the SO data.
• In R built a model using decision tree algorithm we predicted that whether to work on the SO proactively or not.
• If the SO flag with proactive then demand planner give more preference to that particular SO.

Skills

Artificial Intelligence (AI), Keras, Machine Learning, NumPy, Pandas