Birat Poudel

My research interests and professional contributions encompass: 1. Conversational AI Systems: Developing sophisticated dialogue systems and voice-based interfaces (IVR) for healthcare and enterprise applications 2. Model Evaluations: Designing and implementing comprehensive evaluation frameworks for LLMs and ML models, including: a. Automated testing pipelines for model performance metrics b. Bias detection and fairness analysis across different demographic groups c. Robustness testing through adversarial examples and edge cases d. A/B testing frameworks for comparing model versions in production e. LLM-as-Judge evaluation for qualitative assessment of model outputs f. Integration of DeepEval for automated testing and evaluation of LLM applications g. Langfuse for tracing, monitoring and debugging LLM interactions in production 3. Multi-Agent Systems: Designing and implementing scalable AI agent for complex task orchestration 4. Natural Language Processing: Implementing state-of-the-art NLP models for semantic understanding, intent classification, and empathetic response generation 5. Retrieval-Augmented Generation (RAG): Architecting knowledge-grounded systems with significant performance improvements through novel optimization techniques 6. Time Series Analysis & Forecasting: Applying advanced statistical and deep learning methods (SARIMA, LSTM, Prophet, etc.) for predictive modeling