I am Dhiman Deb, working as Software Development Engineer II at Oracle with the Data and Analytics product development team for the last 2 years. My total experience in the industry is 7+ years, spanned across multiple business domains such as Banking & Retail, Oil and Gas, Supply Chain, Telecom, etc.
At Oracle, I work with the product development team to develop new features and functionalities to enable customers to have a better experience in the Supply chain area, starting from Order to Cash, Procurement, Order management, installed base, Maintenance, etc., using various tools such as OBIEE, Oracle Analytics Cloud and technologies such as Machine Learning, Artificial Intelligence, Process Automation using Python, Scala & Spark, etc. We have observed that we can better equip the customer to stock their inventory based on the experience with the supplier and their delivery performance.
We have been using various technologies to prepare a data pipeline and feed the data to analytics applications for Machine learning consumption. Source data is getting injected from multiple source systems such as CSV and Oracle databases into a data warehouse using Oracle Data Integrator (ODI). We have completed various data manipulation steps on source data using ODI inbuilt transformation. Then transformed data got pushed into the Data flow of Oracle Analytics Cloud (OAC), and a pre-built & tested Machine Learning model (Numerical prediction) was used to forecast future demand. Later we visualized the prediction using the OAC data visualization tool.
There are two scenarios particularly: I have applied AI/ML apart from various small/medium initiatives where I have used python to automate or build tools for the organization for better customer experience.
In the first scenario, where I have used the BERT model along with my team members to develop an NLP solution that will provide Oracle-specific answers and links to documents for various analytical and business-related questions.
On the second one, where we are trying to figure out future demand based upon various parameters such as season, past years’ demand, etc.
So far, we have achieved 80% accuracy with the current model, and the entire data pipeline is performing as per expectation. We have also classified suppliers into various segments based on the delivery timeline, such as on-time, late or early.
The whole solution has been tested and demoed to a peer group and published over the Oracle marketplace for the consumption of various Oracle customers.
I have just started Great Learning’s PGP Artificial Intelligence and Machine Learning Course. However, in a very short span, I am able to brush up on my skills as well as get deeper insights from mentors, industry experts, and various study materials.
