Manvik Talwar

Developed Dashboard Stats API aggregating real-time KPIs across student/educator data; adopted as primary analytics endpoint. Built Filtered Submission API with cursor-based pagination, reducing submission list load time by 5×. Implemented trend, completion, and volume analytics APIs with query-level optimizations yielding 40% faster response times. Formulated Submission schema with composite indexes, improving throughput and reliability under high concurrency.

Architected and deployed 'Sofia' — an enterprise RAG chatbot improving internal knowledge retrieval and employee productivity; live on company servers. Engineered semantic search pipeline with custom skillsets for PDF text/image indexing, reducing API response latency by optimizing routing and payload size. Built citation management system with MySQL tagging, enabling clickable source references in every LLM response. Designed prompt engineering strategies to improve response relevance and consistency in LLM-based enterprise workflows. Evaluated LLM performance using latency, accuracy, and response quality metrics to optimize user-facing AI features.