Sri N

Build and maintain Python-based backend APIs for QuickBooks Money, supporting high-volume internal services. Improve service reliability by integrating ML and rule-based validation, reducing user-facing errors. Enhanced monitoring and logging using AWS CloudWatch, cutting issue diagnosis time by ~40%. Support CI/CD pipelines through testing, code reviews, and safe deployments with rollback strategies. Collaborate with Product and QA teams to improve sprint delivery and reduce rework.

Supported applied AI and Data Science projects by converting high-level research questions into Python-based experiments with clear evaluation metrics and documented results. • Worked with real-world datasets (including large transactional and behavioral datasets used in academic projects) to compare traditional models and early LLM-based approaches across accuracy, interpretability, and performance. • Explored vector search and retrieval techniques by building small-scale FAISS and chroma indexes for class and research projects, experimenting with indexing parameters and caching to understand latency and retrieval trade-offs. • Gained hands-on exposure to Retrieval-Augmented Generation (RAG) by prototyping simple retrieval + generation pipeline and evaluating how differently chunking, retrieval and ranking choices are effecting answer quality. • Compared outputs from multiple Language Models (Such as GPT-style models and other foundation models available at the time) to understand differences in response quality, consistency, and failure patterns during controlled experiments. • Studied common LLM failure modes, including hallucinations and inconsistent responses, and experimented with lightweight checks and prompt-level constraints to improve reliability in academic settings. • Built simple FastAPI-based endpoints and demo interfaces to share model outputs and experiment results with faculty and peers, enabling faster feedback and iteration on research ideas. • Developed reusable evaluation scripts to measure accuracy, latency, and basic drift across experiments, helping ensure results were comparable across different runs and models. • Collaborated closely with faculty and student teams, participating in design discussions, experiment reviews and result interpretation while documenting findings for future students to build on.