RECRUITING SERVICE GRAPHICAL USER INTERFACE
A recruiting service is disclosed that generates profiles of software developers having specific skills. Public code repositories are examined to identify projects of software developers. The projects are analyzed to estimate the number of years of experience a software developer has with an individual language and determine a score with respect to other developers. Social media information and a messaging link may also be provided with each profile. A graphical user interface for displaying the information is disclosed.
Latest GILD, Inc. Patents:
This application claims priority under 35 U.S.C. §120 and is a Continuation of co-pending U.S. patent application Ser. No. 13/493,791, filed Jun. 11, 2012, by Bonmassar et al, which claims priority under 35 U.S.C. §119(e) to U.S. Provisional Application No. 61/640,656, filed Apr. 30, 2012, entitled RECRUITING SERVICE GRAPHICAL USER INTERFACE, the contents of each of which are hereby incorporated by reference.
FIELD OF THE INVENTIONThe present invention is generally related to employment recruitment tools. More particularly, the present invention is directed to a user interface, search technology, and scoring technique to automatically provide information to aid in recruiting software developers.
BACKGROUND OF THE INVENTIONRecruiting skilled software developers is a difficult task. How does one find qualified candidates? Many of the conventional recruiting approaches based on reviewing resumes do not work well for recruiting software developers.
One of the problems in the prior art is identifying individuals who have expertise in specific software languages as well as the passion and ingenuity to solve specific problems. How does a recruiter evaluate the actual skills and talents of prospective candidates? Education alone is not adequate to determine actual talent. Nor is the number of years working in industry a good measure of talent.
Another problem in the prior art is identifying whether a candidate for a software development position will be a good social fit for a company. Conventional resumes do not provide a good indicator of the social fit of a candidate.
One aspect of these problems in the prior art is that it is difficult to perform a pre-screening to identify talented candidates to fill a software development position. As a result, many companies waste enormous amounts of time trying to find qualified candidates to fill software development positions. Additionally, the difficulties in assessing the actual talent of a candidate means that companies sometimes end up with employees that cannot perform as expected.
SUMMARY OF THE INVENTIONA recruiting service generates a graphical user interface in response to a query. The graphical user interface provides profile information for software developers. The profile information includes a ranking based on analysis of public code repositories and may also include other information regarding the knowledge, experience, and influence of a developer. The profile information may also be augmented with additional social media information, such as social media links for a developer. A messaging link may also optionally be provided to contact a developer. The recruiting service thus permits a user to input a query to find software developers with specific skills and receive a graphical user interface providing objective evaluation information based on the code written by the developer.
In some situations a large organization could maintain the recruiting service as an in-house tool available to users within the organization via a local area network or Intranet. However, more generally the recruiting service may be implemented as a web-hosted service available over the Internet to individuals, companies, or organizations seeking to obtain information on potential candidates for software development positions.
An exemplary set of Internet information sources is illustrated in
At least one other source of information is preferably accessed to obtain additional information for each profile. Another potential source of information the recruiting service can access are forum and discussion groups 110 used by programmers, such as Stack Overflow (operated by Stack Exchange, Inc. of New York, N.Y.), and news groups like Hacker News (a social news website about hacking and startup companies) or Android developer mailing lists. For example, forum and discussion groups may be used to provide a source of information on the reputation and influence of individual developers. Additionally another option is for the recruiting service to access general or social media sites 115, such as those provided by companies such as Facebook, Inc. and LinkedIn, Inc. Another source of information are contact services and social intelligence services 120, which provide resources to identify individuals from partial contact information and otherwise expand an initial set of contact information into a wider set of contact information and social information from which links to social media can be determined. More generally, other public information sources 125 may also be searched as well that are relevant to determining the influence, skills, or biographical information about software developers, such as professional network sites.
A user utilizes a computer 102 in communication with the recruiting service 100 via the Internet. The user's computer 102 displays a graphical user interface generated by the recruiting service 100. A user searching for candidates to fill a software development position accesses the recruiting service 100 to input a query 130 defining an initial candidate specification, such as proficiency in one or more programming languages. Other examples of a candidate specification include a geographical area specification. In response, the graphical user interface generated by the recruiting service provides a listing of profiles of potential candidates as illustrated by arrow 140, which may also be presented in a ranked order. The user can then request more detailed profile information for individual candidates. An exemplary set of profile information includes the number of years and relative ranking of the candidate in different programming languages, an influence score, overall experience level, a summary of programming projects and links to the projects, a summary of employment history, and social media information. A messaging link is preferably provided to permit the candidate to be contacted either directly via email (e.g., via either anonymous or non-anonymous email) or by other contact modalities (e.g., messaging, phone, etc.).
In this example, a skill query based on “Java” skills is input into search field 205. A search button 207 permits the search to be triggered. This results in an initial listing 220 of profiles 225. In one implementation the profiles are sorted and ranked by overall knowledge. Other profile information may be displayed in the initial listing such as the developer's name 227, photo 229 (if available), brief summary of employment history 231 (if available), and ranked scoring 233 in different programming languages including those in the query and other selected languages for the profile. Thus in this example, the ranked scoring includes the Java language ranking first (because the query was for Java) along with other top scores. The user interface may also provide an indication of the ranking in terms of the top rankings (e.g., through a set of top rankings, such as top 10%, 20%, or 30%) via a tab other visual indicator. Thus, the user can quickly search for profiles in the initial listing corresponding to developers that are knowledgeable and skilled in a language of interest.
The graphical user interface permits a user to select an individual profile and then displays detailed profile information for the individual profile.
The code analysis 310 is based on analyzing code from code repositories to provide objective information regarding a minimum number of years of experience in a particular programming language as well as an objective analysis of the code itself to providing a ranking of the developer's skills. The full analysis includes the estimated number of years of experience with each language. Tabs are provided indicating top rankings (e.g., through a set useful to the end-user, such as top 10%, top 20%, top 30%, etc.). The number of views by others and the adoption of code by others may be used to generate the influence score 325 as a measure of how influential the programmer is.
A bio summary 430 may be extracted from social media. Alternatively, in one embodiment, a software developer is permitted to check their profile and take ownership of their profile in the sense of providing some limited voluntary inputs, such as bio summary information, and also provide feedback on any errors.
A summary of projects 440 accessible in code repositories is also provided. The summary preferably also includes links to the code in the repository for each project 445 for users interested in performing a more detailed analysis of the code itself. Additionally, information about the project may be included such as file size in terms of number of lines of code, number of views by other developers, and number of collaborators.
A crawler 505 is provided to crawl code repositories. For example, the crawler may use an API for code hosting sites such as GitHub. A new candidate profile generation module 510 determines whether the crawler has identified a new developer. If so, a profile ID is generated to build a new profile. A code file type analysis module 515 determined the file type of files being crawled. After the file type has been determined, the language-specific code analysis module is selected by module 520. Scoring and cheating detection is then performed by module 525. Profiles are stored in a profile information database 530. A social media access module 535 provides access to social media information sites and a social media aggregation module 540 correlates aggregated social media information for individual profiles. A messaging interface 545 is included in one embodiment as a means for recruiters to contact individual developers. However it will be understood the messaging interface 545 may be omitted in some implementations. The messaging may, for example, be brokered in the sense of cloaking the user information and email address of the recruiter during initial attempts to contact a developer. A recruiting search engine and graphical user interface module 550 is responsible for generating the graphical user interface that is provided for display on a user's computer.
The new candidate profile generation module 510 utilizes author information from crawled sites to detect that there is a new developer to be added to the system. Code repository sites include author information for each project. This author information is searched by the crawler. Each individual person with a profile has a unique ID. The unique ID is created the first time an individual programmer's name is discovered in crawling author information in code hosting sites. For example, when the crawler finds the names of people that have contributed code to a code hosting site, the system compares the unique ID from the network that the person is found on to the unique IDs in the database of the recruiting service for that network. If an ID doesn't exist, a new user ID is created.
The crawling process generated project information for each developer. One way to obtain project information for a particular person is specifically ask a code hosting site (or content site like Stack Overflow) for a list of projects for each developer. For example, this may be done through an API for sites such as GitHub and Stack Overflow.
The crawling is updated regularly and the profiles are refreshed according to a cycle. An exemplary refresh cycle is a two-week profile refresh cycle. That is to say, the update from the crawlers may be constant, but the profiles may be updated according to a schedule, such as every two weeks.
One aspect of the crawling process is that the source code for a particular project is downloaded for analysis. As illustrative examples, the source may be downloaded using technologies such as Git, SVN, Mercury, and CVS, which are technologies that allow for synchronization with the local computer of a code repository.
It is preferable to download all of the available information in a repository for analysis. However, note that the source code for a project may be in any one of a variety of different file types.
Downloaded files are then processed, starting first with the code file type analysis module 515. An individual file is analyzed to determine what's in it by looking at file extensions and the binary data or text that the file contains. Specific patterns in the source code are analyzed. For example, specific languages—like Ruby—always start with a certain few lines of code—e.g. the hashbang—so that by looking for the patterns of a specific set of keywords in the file permits the language to be identified. Additionally, the analysis of the code can include looking for the “magic number”—a set of bytes at the beginning of the file that indicates file type. For example, images always start with a specific byte configuration. The pattern associated with each different file type is checked until a match is found. The pattern matching may be performed, for example, using a sequence of if-then clauses to identify the file type of a particular file.
After the file type is determined, the language specific code analysis selection module 520 makes a selection of an evaluation tool or tools appropriate for the language of the file type. Most software languages have evaluation tools to evaluate the quality and complexity of the coding. The evaluation tools are specific to a particular language and may, for example, look at the length of the code and patterns in the code. For many cases, the evaluation tools for a specific language are open source and/or available from commercial vendors. For example, there is a unique set of tools to evaluate Ruby—tools that differ from those used to evaluate C++. Thus if the recruiting service is designed to analyze code in languages such as Java, Scala, Shell, ActionScript, XML, CSS, HTML, Groovy, PHP, Perl, Python, Lisp, etc. then the system includes the corresponding evaluation tool for each supported language. Thus, the recruiting service includes a wide range of evaluation tools to support different languages and makes the selection of the proper evaluation tool based on the file type. The file is then analyzed using the appropriate selected tool(s) for the language associated with the file type.
The scoring and cheating detection module 525 utilizes the evaluation of the code and also information from the commit log for the file. To identify the author of the code of a particular file, the commit log is evaluated for the repository. The commit log is a list of who did what for the repository. This permits an evaluation of the developer's specific contribution(s) to that project. For example, by analyzing the commit log an evaluation can be made of the time(s) when the developer made a contribution.
The process is continued for all of the developer's repositories to permit a determination to be made of what languages that a person has experience in and how much experience the person has in each language. Recruiting service downloads all the developer's repositories and evaluates their contributions to determine the languages they've written code in. To evaluate experience the commit log is examined to look at the date and time of when the developer contributed to the project. Different factors can be used to determine actual experience. One factor is that contributions can be evaluated by their frequency and regularity to weight the actual number of years of experience in a particular language.
As an illustrative example, consider a developer named Joe. If Joe started contributing to a project 3 years ago, and the commit log shows that he's been contributing regularly to it, that's an indication that he has 3 years of actual experience. Thus, a weighting function can take into account the frequency and regularity of Joe's contributions. For example, if Joe has been making four or more contributions per year that's an indication that Joe has been regularly working on Java.
However, if Joe made a single contribution 3 years ago and just contributed again for the first time 2 months ago, then the commit log indicates “episodic” contributions with a wide spacing between contributions. For this second case, the weighting factor can be used to reduce Joe's number of years of experience such that he does not get 3 years of experience credit. The exact weighting function chosen can be empirically determined based on common behavior patterns of software developers. For example, if the commit log shows Joe made a single contribution to a Java project 3 years ago and made a second smaller contribution a month ago there could be a possibility that Joe is either 1) trying to “inflate” his resume about the number of years of experience he has in Java; or 2) may have become aware of the recruiting service and is intentionally trying to trick the recruiting service. In this example, the weighting function may also include one or more rules to discount recent contributions, particularly those of a minor character, such as a minor code tweak or a contribution made with many other contributors.
Thus, while the raw data provides an indication of a maximum potential number of years of experience, a weighting function may include different factors related to frequency of contribution, size of contribution, and number of co-contributors to perform a weighting function to arrive at a more accurate interpretation of the number of years of experience for a developer. The weighting function may be determined empirically, based on observations about the way software developers normally work, to optimize different weighting factors and periodically adjusted to discourage gaming of the system. Other types of gaming (such as posting the same code at different times on different sites or plagiarizing code from others) could also, in theory, be checked as part of a larger fraud detection function.
As an illustrative example, patterns in a commit log may be examined for suspicious factors of how the developer is developing his/her source code. The simplest example of a developer cheating is that the developer downloads someone else's source code, opens their own repository, and submits that same exact code to the new repository. In that case, there would be a huge update all at once, and then nothing else in terms of activity. This is inconsistent with normal commit log behavior in which a user normally makes a series of regular contributions over time. An honest developer would normally (except for perhaps extremely small projects) be consistently committing code they are developing for their project. As a result when a huge aberrant spike occurs in a commit log a presumption can be made that there is a high likelihood that cheating has occurred. In this case, the weighting function can severely or totally discount the project, i.e., give it extremely little or no credit.
As previously described, in one embodiment there are three kinds of scores that are calculated for each developer. These include knowledge, experience and influence. This level of scoring provides a variety of useful information to evaluate candidates. However it will be understand that the recruiting service could also be implemented with a subset of this set of scores.
The scores are preferably calculated on a language-specific basis and an overall basis. Language specific scores are useful to evaluate skills in a particular language. However, generating an overall score provides an additional indicator of a developer's talent.
An exemplary language-specific scoring process will now be described. In one implementation, to determine a knowledge score, an examination is made of lines of code and the number of repositories that the developer has contributed to. The score is then calculated by a function that weights the total number of lines of code in all of the different repositories. That is, a developer who has written more lines of code has more experience and credit is given for contributing to different repositories. However, the number of lines of code can reach very large numbers. Thus, one way to score knowledge is apply a logarithm function based on the number of lines of code. As one example, a knowledge score for a developer can be generated using a natural log curve: ln(lines)×number of repositories, where “×” is the multiplication operation and this equation is a simplified equation to illustrate a general approach that one of ordinary skill in the art would further optimize for a particular implementation to optimize empirical results. Other variations based on a logarithm function are also possible and other factors could be included in determining a knowledge score.
An exemplary language-specific scoring process for experience looks at different factors indicative of experience and then weights the factors. For example, to calculate a total experience score the individual experience scores may be combined with work experience, i.e., Total Experience=Individual experience skill experience scores×work experience. Of course, many variations are possible in terms of weighting individual skills experience with work experience. In one implementation the process looks at the lines of code that have been written with particular attention to the lines of code written per day, in addition to commits per day, and the number of days of activity. A weighted function SUM(r) can be used to perform an initial analysis of experience in different skills. The scoring can be further weighted by work experience, resulting in simplified equation to determine a language-specific experience score: SUM(r)×Work Experience, where this equation is a simplified equation to illustrate a general approach that one of ordinary skill in the art would further optimize for a particular implementation to optimize empirical results.
An overall score can be calculating by weighting individual scores. An overall experience score can be determined as follows:
Experience=Individual experience skill experience scores×work experience.
An overall knowledge score can be determined using different weighting approaches. In one approach a logarithm function is used to weight the sum of different knowledge skill scores so that a high score requires the developer to have a wide variety of skills:
Knowledge=ln(SUM(Skill Knowledge)), where this equation is a simplified equation to illustrate a general approach that one of ordinary skill in the art would further optimize for a particular implementation to optimize empirical results.
The influence score is a measure of the developer's influence in the larger developer community. In one embodiment the influence score includes how much the developer's code influences other developers. Additionally the influence score may include the developer's influence in social media. For example, the influence score may include a component based on how a developer's projects have influenced others, based for example on the number of followers, forks, and contributors, which may be determined from data within code hosting repositories. However, an individual developer may have different influence in different languages, which has to be taken into account in determining an overall influence score. Additionally, the developer's influence in social media may also be considered, such as weighting the influence in social media by a weighting function. For example, one measure of influence in social media is a Klout score. Thus an exemplary overall influence score may be determined as follows:
Influence=f(Klout)+ln(SUM(Skill Influence))
where a developer is given credit for influence in different skills may also be given some credit for social media influence and where this equation is a simplified equation to illustrate a general approach that one of ordinary skill in the art would further optimize for a particular implementation to optimize empirical results.
The scoring and weighting functions that are applied are determined empirically to give a desired distribution based, for example by examining what weighting functions give the best real-world results at a particular point in time for recruiters. Thus, for example, the actual constants used as weighting factors and aspects of the weighting functions may be varied based on feedback on the usefulness of the scoring for simulated or actual recruiting efforts.
For example, when calculating knowledge, one approach is to look for extensive experience in several languages. This is because in the real world highly knowledgeable developers have a broad range of experiences to draw upon and are skilled in different languages. Thus, even if a developer has top 10% scores in one or two languages, they cannot get a top 10% overall knowledge score, because that only happens when the developer has top-tier scores in several languages. That is to say, breadth counts.
The public code repositories are crawled on a regular basis. In one embodiment all the raw data obtained from a crawl of the public code repositories is saved, except for the source code. That is, it is preferable to save the information obtained by analyzing the code, in addition to the source log itself. The next time the crawler encounters that repository, the source code is downloaded again, and a “refresh” is made based on new contributions. Additionally, the logs are checked to determine the individual(s) that made the new contribution. Thus, if Michael has a repository with a project, the system will also confirm from the log entries who made any new contributions. Thus if Luca makes a follow-on contribution to Michael's project, the follow-on contribution will be credited to Luca. This cross-checking of which individual made which new contribution to a project is useful to improve accuracy and reliability of the scoring.
The social media access module 535 and the social media aggregation module 540 provide a comprehensive set of social media links for each profile. The author information obtained from public repository sites such as GitHub and Stack Overflow may be incomplete or contain inaccuracies. However typically the author information will include at least an email address and perhaps also a name. This information can then be used to obtain additional social media information using commercial services such as Full Contact, Inc. of Denver, Colo., Fliptop, Inc. of San Francisco, Calif. and Rap Leaf of San Francisco, Calif. Many commercial services check by unique information, like email address, or a hash of the email address (a hash is a unique number generated by an email address. That way, companies can match users by email addresses, but protect their privacy by looking at hash numbers). In one embodiment a search of social media sites is performed of all of the sites listed under Full Contact's set of Social Network Types. From this information profile information identifying the names of developers may be generated along with associated information. For example, work history may also be scraped from social networking sites.
Direct scanning of social media sites is also an option, such as the option of scanning sites such as LinkedIn and Google Plus. However, there's usually not a one-to-one results process. For example, if a developer has a common name, such as “John Smith,” a scan based on their name may turn up more than one hit. To find additional social media links for a particular profile it is thus desirable to look for multiple matching factors (location, title, company, name, etc.), and then calculate the probability that it's a match. If the probability is higher than a certain number, the system automatically merges the profiles. If the probability is less than that threshold, the system sends a notification that there needs to be a manual review process.
Once links to social media are identified for a developer they can be refreshed at a rate slower than other information in the public code repositories. Individuals typically add new social networks infrequently and the URLs of social media sites are generally static.
The graphical user interface discussed in this application includes a set of features that are useful in making recruiting decisions. However, it will be understood that subsets of these features may be used. That is, one of ordinary skill in the art would understand that variations in the graphical user interface to include variations of what has been described are possible.
It will also be understood that the scoring techniques described are exemplary. As software evaluation tools increase in their capabilities it will also be understood that other metrics of coding quality and/or complexity could be utilized.
While the invention has been described in conjunction with specific embodiments, it will be understood that it is not intended to limit the invention to the described embodiments. On the contrary, it is intended to cover alternatives, modifications, and equivalents as may be included within the spirit and scope of the invention as defined by the appended claims. The present invention may be practiced without some or all of these specific details. In addition, well known features may not have been described in detail to avoid unnecessarily obscuring the invention.
In accordance with the present invention, the components, process steps, and/or data structures may be implemented using various types of operating systems, programming languages, computing platforms, computer programs, and/or general purpose machines. Methods and graphical user interfaces of the present invention may also be tangibly embodied as a set of computer instructions stored on a computer readable medium, such as a memory device.
Claims
1. A computer implemented method of providing information for recruiting software developers executed on a computer including a processor, a memory and a network interface, comprising:
- receiving, in the processor, via the network interface, commit logs associated with software projects wherein each commit log identifies software developers which have contributed to the software projects including what lines of software code each of the software developers contributed and when the lines of the software code were contributed;
- selecting, in the processor, a first software developer from among the software developers;
- locating, in the processor, from within the commit logs a plurality of instances where the first software developer has contributed the software code to one or more of the software projects;
- determining, in the processor, how many of the lines of the software code the first software developer has contributed in each of the plurality of instances;
- determining, in the processor, when each of the plurality of instances were contributed;
- determining, in the processor, what programming language each of plurality instances were contributed; and
- based upon how many lines of the software code were contributed in each of the plurality of instances and when each of the plurality of instances were contributed, estimating a number of years of experience of the software developer in at least one programming language.
2. The method of claim 1, further comprising creating a profile associated with the first developer and storing to the profile the number of years of experience in the at least one programming language and identification information associated with the first developer to the profile.
3. The method of claim 2, further comprising, based upon the identification information, retrieving, from a social media site, additional information about the first developer and storing the additional information to the profile.
4. The method of claim 2, wherein the additional information includes one or more selectable links, which, when selected, cause information about one or more instances to be displayed.
5. The method of claim 4, wherein the information about the one or more instances includes the lines of the software code associated with the one or more instances.
6. The method of claim 1, further comprising retrieving first software code associated with one or more of the instances, analyzing the first software code, wherein the number of years of experience of the software developer in the at least one programming language is based upon the analyzing of the first software code.
7. The method of claim 6, further comprising analyzing a quality of the first software code.
8. The method of claim 7, further comprising determining a skill level of the first software developer based the quality of the first software code.
9. The method of claim 1, further comprising determining how many times first software code associated with one or more of the instances has been viewed by other software developers.
10. The method of claim 1, further comprising determining how many times first software code associated with one or more of the instances has been copied and incorporated into one or more additional software projects different from a first software project in which the first software code was used.
11. The method of claim 1, further comprising estimating the number of years of experience of the first software developer in a plurality of programming languages.
12. The method of claim 11, further comprising estimating an overall number of years of experience of the first software developer based upon the number of years of experience estimated in each of the plurality of programming languages.
13. The method of claim 1, further comprising determining a frequency at which the plurality of instances were contributed wherein the number of years of experience of the software developer in at least one programming language is estimated based upon the frequency.
14. The method of claim 13 wherein more or less number of years of experience is attributed to the first software developer based upon the frequency.
15. The method of claim 1, further comprising, determining a large number of lines of the software code where contributed in a short time period and reducing a contribution of the large number of lines of the software code to the number of years of experience which are estimated.
16. The method of claim 1, further comprising estimating the number of years of experience of the software developer in at least one programming language at a first time, receiving updated commit logs, locating, in the processor, from within the updated commit logs one or more of new instances where the first software developer has contributed the software code and estimating, at a second time, the number of years of experience based upon the plurality of instances and the one or more new instances.
17. The method of claim 1, further comprising estimating the number of years of experience for each of a plurality of the software developers.
18. The method of claim 1, further comprising determining a rate at which the lines of the software code were generated wherein the number of years of experience is based upon the rate.
19. The method of claim 1, wherein the commit logs are retrieved from one or more public code repositories.
20. The method of claim 1, wherein new commits associated with the commit logs are retrieved on a periodic basis and the number of years of experience is updated on the periodic basis.
21. The method of claim 1, further comprising determining a number of collaborators on a first software project associated with a first instance wherein the number of years of experience which is estimated is based upon the number of collaborators.
Type: Application
Filed: Mar 25, 2014
Publication Date: Jul 24, 2014
Applicant: GILD, Inc. (San Francisco, CA)
Inventors: Luca BONMASSAR (Marina di Massa), John Dane Smilanick (Santa Clara, CA)
Application Number: 14/225,405
International Classification: G06Q 10/10 (20060101);
