Prankur Joshi

Advanced Information Mining (AIM) on SWIFT messages, which have contextual information shared through free text tags by banks to sellers.

• Builded a solution which enabled the customer to reduce the human effort which extracted the Key Information into structured format which could be fed into down streams applications.
• Writing python script to feed prompt to LLM model to extract entities from documents based on the user's query.
• Developed QnA chatbot using Streamlit and achieved 92%+ accuracy across all the information extracted from messages.
• Developed Policy Assistant chatbot for the internal Blazeclan users. This CB is intended for all the employees in the org to have a kiosk for asking any query related to Organization policies.

The customer needs to provide their client with contextual, relevant, and real-time invoice details (total amount, outstanding, payment history) through WhatsApp, leveraging integration with the BigQuery database.

• Created a solution that leverages a GenAI approach on Google Cloud Platform (GCP), utilizing Gemini LLM and Vertex AI Agent Builder.
• Accessed structured data from BigQuery and unstructured data from PDF documents to provide relevant and contextual answers.
• Integrated this solution seamlessly with WhatsApp for text messaging and phone calls for voice queries via the Dialogflow API.

• Clients can identify risk and mitigations in business activities under fair business practices outlined in FINRA (Financial Industry Regulation Authority) and CFR (Code of Federal Regulations) regulations.
• A solution is proposed to address vital risk identification using a combination of ML classifiers such as Naïve Bayes, BERT, and a cosine similarity-based approach in identifying contextual risk.
• Create the dataset which contains 2 features: Risk and Content, using LRR name of Risk Mapping sheet and performing Exploratory Data Analysis (EDA) on Risk mapping data, which includes extracting relevant key phrases and synonyms from each risk class.
• Finding features from the content and semantic similarity scores within the content to find the closer risks and calculate % relevancy score for predicted risks.
• Provide the provision to the user to override the predicted risk, and that updated risk can be appended to the training dataset for CML pipeline and exporting Training dataset to Database hosted on Azure.
• To curate global regulatory laws and rules from official and external public legal sources for client KPMG.
• Reading JSON Files from Microsoft Storage Blob using Service Bus Explorer and writing generic code to fetch Common ID, Headings, and Contents from different structures of JSON CMM files of various sources such as the US, the UK, France, Germany, and India.
• Cleaning data, which includes tokenization, stop words removal, removing punctuation, foreign language characters, and special characters.
• Applying the BERT algorithm to unsupervised data to detect the relevant topics for each section of contents.
• Appending topics to the JSON files, then ingesting the data to the Elasticsearch database.
• Working closely with the DevOps team to deploy the model, handling environment variables, Docker files, and other dependency files, and maintaining the Dev, QA, and Prod environments in Azure.

• Assisting Sales Representatives and Sales Order Specialist to resolve error in a runtime when they sell BMC products during the absence of Salesforce Consultant
• This application provides the best possible resolution of the error to the salesforce users
• Retrieved data from BMC Internal Remedy tool and collected their respective resolutions from respective application teams

• Real Time Signal provides consultation and design services for developing software right from initial product idea for web application
• Gathered requirements from business
• Performed preliminary data analysis using descriptive statistics and handled anomalies such as removing duplicates and missing values
• Applied various Machine Learning algorithms such as Logistic Regression, KNN, Decision Tree Classifier, SVM, ANN

• Responsible for Monitoring and Maintaining Production & Application Servers and Databases
• Coordinating with Database, Unix, Linux, Application, Wintel, Network and Incident Management Teams to resolve the issues and Finding out the Root Cause Analysis