Akshay Chauhan

Logfile Recommender Service, DynaResponse

• Built and scaled NLP-driven support automation for SAP HANA, BI, and Analytics Cloud, cutting incident resolution time by 40% across 500+ customers.
• Designed and delivered LLM and RAG-based autonomous systems for intelligent log analysis and SAP Note recommendations; led LLM fine-tuning POC for domain classification.
• Improved recommendation performance by 40% using Neural Collaborative Filtering and advanced ML techniques.
• Developed, optimized, and deployed production-grade ML models (XGBoost, LightGBM, Word2Vec, Sentence Transformers), maintaining 90%+ code coverage and compliance.

Property Insight Engine

• Built a Property Insight Engine to improve risk assessment, underwriting accuracy, and loss ratios for Property Insurance (LOB) using computer vision and GenAI.
• Trained segmentation and multi-class classification models for roof types, attributes, and damage detection using PyTorch, Detectron2, and OpenCV.
• Developed and deployed a GPT-4 + LangChain–powered chatbot to retrieve and summarize property data from SQL databases for underwriters.
• Implemented scalable data storage and retrieval using MongoDB and MySQL, enabling faster property evaluation and improved customer experience.

Ad Campaign Analytics

• Built an ad campaign analytics platform to uncover customer preference patterns and optimize targeted advertising using NLP and Transformers.
• Engineered high-impact features from historical ad and video data, driving CTR improvement in new campaigns.
• Developed predictive models (Random Forest, NLP-based models) to forecast ad campaign performance prior to launch.
• Generated new ad copy automatically using GAN-based text generation and GPT, informed by insights from past campaign performance.
• Integrated Google Vision API and GCP to extract visual and contextual features from ad creatives at scale.

Kicker-Ace

• Hands-on experience in computer vision and video analytics, including YOLO-based detection, Kalman Filter tracking, and OpenCV pipelines for sports analytics applications.