At IBM, I spearheaded a retrieval augmented generation assistant for an insurer, indexing policy manuals with LangChain and vector embeddings. I engineered prompts and guardrails that cut hallucinated answers by 31% before release.
I also fine-tuned Hugging Face transformer models on support transcripts, reaching 92% macro F1 for ticket-intent classification. For the assistant rollout, I designed an A/B testing framework that validated a 14% rise in resolution rate.
At Mphasis, I migrated legacy SQL Server workloads to Snowflake for a retail banking client and orchestrated Airflow ELT pipelines that shrank batch windows by 65%. I also improved data accuracy to 99.5% with data quality checks.
At Cybage Software, I analyzed e-commerce purchase patterns and found checkout abandonment behind declining revenue. I built Tableau KPI dashboards and applied regression and A/B testing to support demand planning and checkout improvements.

