Shipped · Product Management · Interview Kickstart
How we built the FAANG Resume Analyzer at Interview Kickstart
Designing and launching a proprietary AI-powered resume analyzer that leveraged unique data to create a high-volume, high-ROI lead generation channel.
- Product Strategy
- Generative AI
- Case Study
- Interview Kickstart
Problem
IK's best asset was thousands of FAANG-placed resumes, but acquisition still leaned on a high-friction webinar. Generic resume tools ignored that data, so site visitors got weak feedback and IK had no low-friction lead engine on the marketing site.
Metrics
- Fine-tuned gpt-3.5-turbo resume analyzer; 8% CAC reduction over 3 months
- ~7-8k successful FAANG-placed resumes as the training set (from ~25k learner resumes)
- Target analysis time under 10 seconds; ~$10k build and ~$40k annual run cost
Decisions
- Train on IK's FAANG-success corpus instead of a generic LLM resume prompt.
- Host models internally with PII masking rather than sending resumes to a public API.
- Pick Gemma 7B on cost-per-inference vs scoring quality; validate with a 10% traffic test.
Outcomes
- Shipped an internally hosted analyzer with PII masking, OCR, and a scored report UI.
- GTM via an intent pop-up, then a 10% traffic A/B test before scaling.
- Turned proprietary resume data into a lead channel instead of another webinar ask.
Longer write-up
Optional long-read with extra context and narrative lives on Notion.