- Delivered technical guidance on algorithm design, complexity analysis, and debugging in Python-based computational assignments.
ChatSprout 2.0 – LLM-powered English Training Assistant
Goal
Help non-native English professionals improve workplace communication (networking, meetings, and feedback) with real-time suggestions.
Approach
Built an LLM assistant with a RAG pipeline (embedding → retrieval → generation), plus a low-similarity fallback to reduce hallucinations.
Results
Improved Top-3 scenario match accuracy to ~80% and kept average response time at 1.5–2.0 seconds via prompt + API strategy optimization.