Naresh Bodicherla

Document intelligence:-

Developed/deployed a tool for extraction of relevant info from the document, making it easy for underwriters to make decisions and understanding the risk levels.

• Technology Stack: Pymupdf, Layout llmv2, Azure form recognizer, Azure Cognitive search, Azure OCR, Question Answering model, Flask API, Docker,Multi-threading, Multi-processing.
• Key Achievements: This Capability/tool has been converted into a reusable asset which can be used in Future Projects.

Video Similarity:-

Led a team with primary task of building a framework to index and recommendation of similar videos based on objects identified in video and Audio transcripts.

• Technology Stack: Katna, YOLO, Vision Transformers, Nearest Neighbours, Wave2vec, Azure Cognitive Search, Flask
• Key Achievements: A Novel approach to index videos to be used for video analytics.

Subrogation Classification / Prediction:-

Prediction of Subrogation in property claims sending a warning flag for claims handler while handling multiple languages.

• Technology Stack: NLP, Pyspark, Azure Data Bricks, Sentiment Analysis, Multiple Languages handling , Ensemble models.
• Key Achievements: Increased percentage of subrogation claims identification by 40%.

Medical Invoice Analysis:-

Analyzing and indexing medical invoices to integrate into the various models used during claim cycle and reducing the human intervention required.

• Technology Stack: NLP, Pyspark, CascadeTabNet(table detection algorithm),
• Key Achievements: Reduced Manual processing of medical claims by 60 %.

Weather Anomaly Correlation:-

Verifying the correlation between claims lodged for property damage and weather anamolies.

• Technology Stack: Pyspark, Statistical Analysys T Test, Anamoly detection.
• Key Achievements: Identification of new risk area for property Insurance thereby reducing the impact it will have on future policies

Property claim Quick Settlement :-

Built a ML model using NLP, Random forest and XG Boost to classify property claim quick process or not

· Technology Stack: Python, NLTK, TextBlob,NumPy, SciKit-Learn, Flask

· Used ADASYN to address imbalanced dataset problem

Key Achievements : Achieved profit gains of 10%.

Vehicle loss Prediction:-

Built a ML model to predict the loss amount prediction at the FNOL (fist notification of loss) by insurer or third party and identifying the key factors

• Technology Stack: Python, NLTK, TextBlob,NumPy, SciKit-Learn, Flask,PCA,SVD,XGboost,Grid search
• Key Achievements: Identification of high net loss claims and increasing the scrutiny over those claims to reduce the loss.

• maintaining Oracle/ SQL databases across cloud / on premise
• Extraction and transformation of data from databases.