Anil Kumar Reddy

End-to-End Model Development and Deployment: Designed, built, and deployed ML models to process 6-8 million documents daily for ESG Risk Compliance (ERC) product. Developed and integrated models for entity extraction, topic extraction, sentiment analysis, theme detection, summarization, noise classification, severity detection, and deduplication at scale. Fine-tuned and deployed GPT-3.5, GPT-4o, and transformer-based models for specialized tasks. Created a custom Retrieval-Augmented Generation (RAG) pipeline to extract factual information from large datasets. Infrastructure and Scalability: Set up scalable infrastructure for vector databases (Qdrant) and engine pipeline, ensuring high performance and reliability. Deployed models using Docker and Kubernetes, implementing containerized solutions for production environments. Operational Excellence: Established robust test harnesses to validate models before production deployment, improving reliability and performance. Managed on-call engineering responsibilities, ensuring smooth operation and quick resolution of pipeline issues.

Led a team of 2 to develop Speech-analytics tool to extract the insights from conversations. Developed Aspect Based Sentiment Classification model for Survey analytics pipeline.

Researched on multilingual translational and transliteration models. Optimized the in-house built transliterated model and reduced the latencies 3-6x.

Designed and developed AI models to extract information from various BFSI documents. Implemented Question Answering and rule based models to extract various KPIs from Annual reports for a global ESG Analytics company. Designed and Implemented job processing engine using Celery, RabbitMQ, Redis and helped scaling the throughput. Set-up AI infrastructure in the organization and played a key role in adding intelligence to the migration product. Content analysis with AI/ML(Auto classification, Information extraction, Clustering). Implemented ROT(Redundant Obsolete Trivial) module by finding the duplicates and near duplicates on both text and image data. Implemented PII(personally Identifiable Information) and redacted the information for compliance purpose.