Cool project.<p>After a cursory look through several of the top lines in different industries, it looks like your system strongly favors highly specified descriptions combined with positive numerical growth. For example:<p><i>Decreased uninstallation rate by 40% by introducing an interactive tutorial at app launch</i> (Product Management)<p><i>Analyzed industry trends in the automobile sector and presented long and short equity investment ideas for 12 large-cap stocks that outperformed the Bloomberg sector benchmark by 7% in 2014.</i> (Trading)<p>In other less common cases, it looks like your system selects underspecified lines, like the following:<p><i>Created a DCF valuation model to analyze a potential IPO of a major technology startup in New York</i> (Quantitative Analysis)<p>...which is interesting to me, because that suggests this is happening manually right now. So your database is strongly predisposed to be an aid in candidate searches optimized for specific metrics (like trading, management and data science). On the other hand, your system is probably not all that helpful for things like technical information security, where metrics are much more difficult to judge and specify.<p>I'm interested in how you're doing this, because companies like LinkedIn are obviously very defensive against people crawling their resumes. But to do this more efficiently (and I'd say accurately too), you'd probably want to have a crawler manually weighted towards the top <i>n</i> companies in each target industry, with an NLP system recognizing the salient points of employees' resumes who work at those companies. Is that something you're working on or plan to work on, or are you going to do the resume and line curation manually?<p>As another point, I'd challenge your priors a bit. I don't know that you have a strong value proposition with just lines and no other specific context. What might be more helpful is the following:<p>1. Find a way to add the structural context of the resume instead of just salient lines: did this line appear under a job description? Was it under an accomplishments heading?<p>2. What if, instead of the most impressive lines, you develop this out to analyze the entire resume as a product? For example, collect as many resumes as you can, break these out into a universe of features, then produce statistics and visualizations on how close a resume is to optimal for a particular company. "88% of engineers at Google have this length, with these headings, etc".<p>Optimize the long tail of metrics that can be quickly changed for applicants, not the high impact permanent ones (like how long they've been at each job, which university they attended, etc).<p>3. Are you sure you want to target this product directly to applicants? If you develop this out into an effective data analytics product for how optimal a resume is for a specific company or likely it is to receive an interview, you could produce something targeted for recruiters that is worth a lot more.<p>In my opinion, this is a great first step, but you could be building a novel approach to recruiting here, targeted towards recruiters and companies instead of candidates.