Project narrative
SDRs struggle to write personalized, prospect-aware emails at scale. Magic Mail automates this by generating tailored outreach and objection responses using fine-tuned LLMs.
I analyzed existing high-performing sales emails to identify effective structures, personalization patterns, and objection-handling strategies. These insights guided prompt engineering experiments and the curation of a fine-tuning dataset covering prospect details, industry context, and common objections.
The fine-tuned model generates both initial outreach and contextual replies. Magic Mail reduced email composition time by 250% and improved positive reply rates by 80% across pilot customers, freeing SDRs to focus on relationship-building.