Muthukumar Rajendran

• Led the development of agentic AI solutions focused on risk, compliance, and regulatory frameworks. • Conducted research on fine-tuning large language models to enhance machine learning capabilities. • Collaborated with cross-functional teams to ensure alignment with regulatory standards and business objectives.

- Architected an automated ML engineering platform using GPT models for code generation, testing, and optimization - cutting deployment time by 40%. - Developed a hybrid product recommendation system (Graph Neural Network + LLAMA 3.3, fine-tuned with LoRA) achieving 92% accuracy and real-time inference for sales enablement. - Built Critical Risk Factors Q&A multi agentic system using LangGraph; delivered 84% accuracy and real-time interpretability for underwriter queries. -Engineered a web intelligence framework (LLAMA 3.3 + BERT) improving entity resolution accuracy by 9% over vendor baselines and consolidating multiple APIs into a unified endpoint. - Developed and containerized NLP pipelines for business website optimization, achieving 92% precision and reducing manual review costs by half. - Deployed ML solutions on Azure ML + Docker + FastAPI, ensuring scalability, observability, and CI/CD integration for enterprise adoption.

- Led design of data-driven risk analytics solutions, integrating predictive modeling and LLMs to enhance underwriting intelligence and reduce operational latency by over 50%. - Built industry-specific clustering and knowledge-graph frameworks for entity relationship analysis and risk factor detection, improving model accuracy to 89%+. - Created intelligent feature engineering libraries in Python, accelerating data preprocessing and boosting model reproducibility across multiple risk domains. - Delivered interactive analytics dashboards (Plotly/Dash/Streamlit) enabling underwriters and business leaders to visualize insights and identify emerging risk trends. - Designed experimentation pipelines for model evaluation and drift monitoring to ensure compliance and interpretability in high-impact financial decisions. - Collaborated cross-functionally with product and underwriting teams to transform analytical findings into deployable, business-ready insights. - Mentored junior data scientists on advanced modeling (graph analytics, NLP, and unsupervised learning) and best practices in ML experimentation.

- Researched semantic text similarity using deep learning with a cross-functional team. - Analyzed Canadian mental health trends during 2021 using language models. - Developed a novel interactive topic model using DistilBERT and HDBSCAN. - Developed a high-recall RoBERTa model for classifying imbalanced datasets.

- Built and deployed scalable ML models in production, including hybrid recommendation systems, sales forecasting, and pricing optimization pipelines. - Developed time-series and regression models (linear/logistic) for sales prediction and business insights. Designed ML-driven pricing strategies and created interactive dashboards for business and product analytics. - Designed and built knowledge graph–based systems to capture entity relationships and enhance downstream analytics and recommendations. - Developed and deployed transformer-based NLP models for text classification, semantic search, and content understanding in production environments.

Research on deep learning, machine learning and large language models