Data Scientist with 4+ years of commercial experience in analytics,
* LinkedIn Profile
* HackerRank Profile
* Medium Profile
Entity Extraction.
Natural Language Processing
Goal:
The Aim to develop a model that identifies all Key information extraction on medical documents and identity patient name for extracting all information about patient from database.
Roles & Responsibilities:
Created Chatbot for answering all the questions raised by end users about Niyama Digital Health care using OpenAI GPT4. Previously Traditional chatbot was used (RASA) which can answer only fixed questions.
Benefits: The project resulted in a chatbot that not only answered questions efficiently but also provided users with a personalized and interactive experience, ultimately increasing user retention and satisfaction.
Scalability: Designed the chatbot to easily scale as the user base grew, ensuring that it continued to save time and money as the website’s popularity increased
Measurable Impact: The project’s implementation had a measurable impact on reducing support response times and operational costs, ultimately increasing user satisfaction and engagement.
Automatic PHI Form Filling.
Natural Language Processing
Goal:
Developed a cutting-edge Automatic PHI (Protected Health Information) form filling system using OpenAI GPT-4, by creating an Automated Form Filling system. This system resulted in a remarkable 80% time savings for doctors compared to the previous manual data entry process.
Roles & Responsibilities:
AI Solution Development: Led the development of the AI solution, defining the technical requirements and selecting the technology stack, with a primary focus on OpenAI GPT-4 integration.
Natural Language Processing: Developed and fine-tuned the NLP component of the system, ensuring it accurately understood and processed patient responses.
Time Efficiency: With the automatic PHI form filling system in place, doctors no longer needed to transcribe patient responses or manually input data, saving a substantial amount of time during and after patient calls.
Cost Savings: The project resulted in significant cost savings by reducing the administrative workload of doctors, enabling them to concentrate on patient care.
Scalability: Designed the system to effortlessly scale as the volume of doctor-patient calls increased, ensuring that the time and cost savings continued to grow.
Wafer Sensor/Semiconductor Fault Detection.
Machine Learning
Goal:
The Aim to develop a framework for semiconductor faults detection and classification (FDC) to monitor and analyze wafer fabrication profile data from many related process variables to Identify the cause of the faults and thus helps reduce abnormal yield loss.
Roles & Responsibilities:
Analyze the stakeholder requirements,Performed Feature Engineering on data having 592 columns using Python libraries like Pandas, Numpy, Matplotlib.
Segmented data by creating Clusters using K-Means.
Built and trained a model using Random Forest Classification and XGBoost Classification.
Maintained the code changes in Bitbucket and Deployed using MLOPS (DVC and MLFlow) for model mon and utilized - Circleci for performing cicd pipeline and deployed in - AWS EC2.
shredder Machine Hand Detection.
Goal:
The Main Goal of the project is to detect hand of a person before it gets chopped by the Shredder Machine in WAREHOUSE. If the hand crosses the ALERT line an ALERT sound must be given, and if the hand crosses the SAFETY line, the shredder machine must be made shut down.
Roles & Responsibilities:
Analyze the stakeholder requirements, Performed Data Collection on different lighting Conditions and annotated 4000+ images using LableImg.
Used SSD with Mobilenet for hand detection
Raspberry pi Micro Controller is used to Shut the Machine if the object crosses the Safety Line.
On Premises setup was designed to Deploy the Project.