Line Items & Transactions ExtractionLeveraged Azure Document Intelligence OCR - New Feature development - extracting line items from invoices & transactions from bank statements - Python - Algorithm creationML models Implementation - US DocumentsElixir Implementation - Debugging - Feature Updates of the model - document OCR analysisPython to Elixir PortingPorted Python-functionalities to Elixir - contribution to a more scalable and robust code - DebuggingAI Auto Fill - Feature CreationPython - Elixir - Debugging - Stabilizing the Prompt and the Workflow - OpenAI - Prompt Engineering - TestingIntroducing Custom Templates - Lookup Feature ExtensionDesigned and implemented customized Audit document templates for clients.Ensured precision in data extraction - Improved the overall audit process.Extract Key value pairs - by anchor - Python - Elixir - TestingOptimizing Lookup - Existing FeatureFixed the deviation in Key-Value PairsImplemented Multi-Line Key Detection - Python - OCR - Algorithm CreationOptimized the Lookup - Location & Key-Value based - Python - Testing

Advanced Recommendation System for PatientsDeveloped an advanced patient recommendation system - accuracy 90% - above.Utilized LangChain, RAG, GPT4 models - Tables extraction from PDFs with Camelot.Employed NLP Prompt Engineering - refine patients’ recommendations - data analysis.Integrated InstructEmbeddings - HuggingFace Transformers - enhanced semantic understanding.Deployed via StreamLit - intuitive UI - GoogleColab - collaborative development.

Contract REview and Playbook Creation - Legal documentsDeveloped an improved App - two-in-one functionality - Contract REview with the created Playbook - Prompt Engineering- Camelot - RAG - NLP - GPT4Document Due Diligence and API DevelopmentDue Diligence conduct - Integrated GPT3.5 Turbo - Prompt Engineering - NLPAPI development - Python - maintaining robustness & security - code refactoringDocument Retrieval & Similarity - LLM Chains Utilized Langchain - LLMs Chains - RAG - Vector Embeddings - ElasticSearch - OpenAIInformation Redaction and other implementationsAccomplished redaction - Spacy - sensitive information - NLP - PyMuPDF - PythonDeveloped Atticus AI - MS WORD Plugin’s Backend - Python Upskilled - Learnt JavaScript - aiming for React.jsWorking with MongoDB and Optimization with AsyncIntegrated MongoDB for efficient storage & retrieval of data - Draft ChatImplemented Asyncio - Python - enhance responsiveness & scalability - Contract ReviewOpenAI API Assistants and PostmanCreated Contract Review API - OpenAI API Assistant - Developed Postman API & Curl Commands

Form Filling - GPT InteractionImplemented form-filling functionality - PyPDF2 - Integrated GPT4 - 80-90% accuracy - NLPUtilize User Comments’ - Drafts’ Improvisation Developed Drafts’ Improvisation feature - functionality for improving drafts using User’s CommentsIntegrated GPT4 - PromptEngineering - Dual-Feature - Sectional & Entire Draft Level ChangesNext Best 5 Action Prediction through Draft ChatGPT4 - predictive model - next best five action - user interactionNLP - Prompt Engineering - refine & improve Prediction Accuracy through Draft Chat

INSIGHTS – WORLD BANK IBRD DATASET Feb., 2022Data Exploration - K–Means Clustering, XGBoost Regression and Classification. LOAN STATUS & INTEREST RATE PREDICTION - Lending Club Dataset Mar.-Apr., 2022Data Visualization and Predictive Modeling - Xgboost Classifier & Regressor Deployment - Pickle files - Streamlit PythonSIGNATURE & MULTIPLE OBJECT DETECTION - Tobacco800 & cheque custom datasetMay - Jun., 2022 Data Collection and YOLOv5 formatting - YOLO custom object detection - YOLOv5IMAGE ENHANCEMENT & LOGO DETECTION - Custom Dataset Jul., 2022 Threshold & Template Matching using Opencv-Python, Computer Vision