Ganesh Samanta

Experience: - 8 years and 8 months

For the last 5 years and 8 months, I have been an integral part of the Data Science team at Ericsson, before that I worked for Capgemini as a data scientist for 3 years. I have demonstrated an ability to perform sufficiently, consistently, and repeatedly the whole range of functions associated with the post of Data Scientist.

Certified on Neural Networks and Deep Learning https://www.coursera.org/account/accomplishments/verify/NLUYMGD9FLNQ

Machine learning signature verification in Ericsson

https://www.ericsson.com/en/blog/2022/10/machine-learning-sourcing-contract-compliance

• Date of Birth: 01/01/1990
• Gender: Male
• Nationality: Indian
• Marital Status: Married

• SSD- Validate Signature for sourcing compliance (Sourcing, Ericsson)
• 4SIGHT- Forecasting tool along with self-explanatory analysis (Strategic Finance, Ericsson)
• Darwin- Generate score to install modern machine at outlet (Capgemini)

• Power Award Q3 2022: Rewarded for outstanding contribution to Signature validation and contract readability (Ericsson: Nov,2022)
• Ace Award Q4 2020: 4SIGHT Product: Forecasting self-servicing product, Long-term engagement (Ericsson: Mar,2021)
• Spot Award - Finance External Data Q3 2020: Rewarded for External data inclusion in financial KPI forecast (Ericsson: Oct,2020)
• Mathematics Topper in Higher Secondary board exam 2008: Awarded with "SHRI SATISH CHANDRA AWARD" Prize for securing Highest marks in Mathematics.
• STAR (Special Thanks and Recognition): Outstanding performance and lasting contribution to Insights & Data (I&D), Capgemini (Capgemini: Apr,2017)

• Amlan Manna (Head of Data Science Chapter, Ericsson, Bangalore, India)
• Saktipada Maity (Director of Data Science & Analytics, Capgemini, India)
• Prof. Prasun Das (SQC & OR Division, Indian Statistical Institute, Kolkata, India)

Signature Verification: Enhance sourcing with automated compliance:-

Objective:- Determine potential noncompliance with sourcing contracts by verifying available signatures in contracts/agreement documents.

Development: -Integrated solution to automatically detect and validate signatures on any sourcing contract document to reduce manual work and human intervention.

The developed solution should have a high accuracy in the identification of the signatures.

—Identify the signature page: which pages contain a signature? Technique: NLP classification.

—Locate signature: How many signatures are present on a given page and where are they? Technique: Object detection using YOLO.

—Identify and validate the signatures: Who has signed? Does he/she have the power of attorney from Ericsson to sign? Technique: GAN and CNN

Production: -Training and validation performed on AWS Sagemaker instance with labeled data. Fitted the best model in the AWS training account and deployed the model pipeline in AWS Cloud formation through AWS endpoints for model and lambda functions for triggered interconnected processes.

ML based Hybrid Forecast:- 

Objective:- Assessed financial KPI forecast model driven by external factors.

Designed and developed a machine learning-based solution providing forecast accuracy of ~90% on average compared to ~80% in the existing one for group/market area, assessed hybrid forecast model along with different external factors.

—Automated Forecast app 4SIGHT: Which dockerize and deployed in Azure.

—The process behind 4SIGHT is based on four modules.

1. (Explore):- Visually inspect the data it will base its forecasting on. This data can be collected directly from enterprise systems in the organization and consists of information regarding contracts and payment milestones.

2. (Optimize):- 4SIGHT optimizes by identifying the best configuration in choosing between four different Machine Learning models.

3. (Generalize):- Iterates forecasts for up to three months using different configurations. The user can apply specific configurations to match their preferences or use the pre-set configuration mode.

4. (Forecast):- Generate forecasts including one or several financial variables

Predictive Maintenance: German Leading Automotive Manufacturer:-

—Explored various techniques for Predictive Maintenance through research papers and came up with the feasibility and limitation of techniques and then decided to use logistic regression with clustering.

—Performed exploratory analysis on the robot's activity +behaviors data sets to understand the hidden pattern and insights.

—Problem with production rate, predict preventive messages and maintain the robot in advance to increase production. Clustering the robots on activity +behaviors, logistic regression assessed by suitable probability cut-off value, build multiple linear regression models to score -will tell robot will fail or not. Build R-shiny demo for UI visualization and presentation for internal and external uses. Tools: R, Hive.

Install Modern Machine & promotion of goods at outlet: Consumer Goods Industry:- 

Objective:- 1. Predicting Score to install Modern Machine at outlet  2. Deciding the location & promotion of goods inside a store & investigating such differences useful insights which will improve sales.

—Performed logistic regression to identify the characteristics (Some of the Key Performance Index created from available data) of retail outlet leads to installing the modern machine.

—Association (Market Basket Analysis) among product and chances of purchase a product along customers decided purchased a product. Used segments likelihood of purchase behavior in promotion design for future sales.

—Build in R, show results in R shiny dashboard

Recommendation search engine by Text Mining:-

—Objective was recommended files which are relevant to a query.

—Build a recommendation search engine in Python with the help Natural Language Tool Kit, recommend files which are relevant to a query, retrieve info from different kind of documents, tokenize, stemming, classification and word2vec model based on NLTK corpus

Optimize cost of network flow with different constraints: Manufacturing plant:-

—Created automation R code which creates a matrix for objective, constraint, and sign of optimization problem.

—Find optimize solution by RGLPK and analyze the output like feasibility, conflicts.

—Build in R & R-Shiny Dashboard to show feasibility and optimized path.