At Outlier AI, I design and maintain reusable prompt templates across customer service, legal, marketing, and finance. I built RAG-integrated workflows in LangChain and LlamaIndex for grounded enterprise knowledge base responses.
I reduced LLM hallucinations through structured chain-of-thought and constrained-output prompting. I also led A/B testing of prompt variants, improving task completion quality scores by 40%.
At DigitalEdge Media, I engineered multi-step prompt pipelines for blog, ad copy, and social media generation. They reduced content production time by 60% while maintaining brand consistency across 12 client accounts.
At Boston city college AI Lab, I evaluated prompt sensitivity across GPT-3 variants for NLP benchmarks and documented prompt failure patterns. I contributed to an internal report on few-shot learning effectiveness.

