Nirmal Kuiry

Designed and deployed demand forecasting for QCom and ECom datasets using Python (TensorFlow, Polars), NLP and ML models, improving forecast accuracy by 10-12% and enabling demand planning for clients. Built automated forecasting evaluation pipeline covering 14+ models, reducing manual analysis time by 25%. Deployed LangGraph/LangChain-based multi-agent AI systems with agent-to-agent workflows, memory management, and API integrations (FastAPI, AWS Bedrock).

Results-driven professional with a strong foundation in ML, achieving a 15% improvement in process efficiency. Delivered end-to-end data solutions using Python, LLM that reduced manual reporting effort by 30%, improved retention by 12%, and boosted data accuracy by 18% through automation and data validation.

Data-driven statistician with experience delivering analytical solutions to a worldwide learner base, resolving complex quantitative problems with 98% solution accuracy. Built Power BI dashboards for monitoring resolution metrics. Enhanced query resolution quality by 25% through structured statistical frameworks.

Developed LLM interface APIs with dynamic model routing (GPT-4 class, GPT-3.5) and a RAG pipeline using vector search and reranking. Delivered production-ready, cost-efficient AI agent system. Implemented SFT and RLHF pipelines for multimodal LLM architectures, improving model response accuracy by 9%. Increased predictive model performance by 5-7% through feature engineering and hyperparameter tuning.

Reduced clarification rework by 30% and maintained 87% academic compliance, as assessed through quality audits, repeat-engagement metrics, academic integrity guidelines, and structured review processes.