A Job Finder Chatbot-Based Web Platform: A Use Case for Software

Engineers

Panagiotis Fotiadis, Georgia M. Kapitsaki

and Maria Papoutsoglou

Department of Computer Science, University of Cyprus, Aglantzia, Cyprus

Keywords:

Job Adverts, Chatbot, Hard Skills, Soft Skills, Software Engineering Positions, GitHub.

Abstract:

Finding a job requires browsing through a vast number of position openings, usually at various online sites.

This process becomes more complicated for users when they are considering different potential job locations.

Even though existing tools that automate the process exist (e.g. sonara.ai, dream.jobs), they lack the element

of interactivity with the user and the integration of external resources that contain information on the skills of

the user. With the aim to bridge the gap between job seekers and potential employers by matching resumes

and job listings more efﬁciently and effectively, in this work we are introducing an AI-driven chatbot-based

web platform to assist job seeking. In the initial implementation, we are considering the job seeking needs of

software engineers but more disciplines can be easily added. We have integrated the developer’s CV and the

user’s GitHub account, and are describing the design and implementation process of the web platform, while a

small scale user evaluation has also been performed. We argue that the chatbot can be used as a starting point

for similar job seeking assistants.

1 INTRODUCTION

There is a noticeable disconnect in the current job

market, where employers struggle to ﬁnd candidates

with the right mix of skills and competences, while

job seekers often ﬁnd it challenging to locate op-

portunities that match their speciﬁc skill set, experi-

ence and career aspirations (Aluas et al., 2024). The

essence of personalized job recommendations lies

in their ability to signiﬁcantly increase employment

chances and simplify the job seeking process. By

delivering tailored job suggestions that align closely

with an individual’s skills, experience, and career

aspirations, these systems transform the task of job

hunting into a more manageable and targeted en-

deavor. The utilization of Artiﬁcial Intelligence (AI)

and machine learning technologies presents an op-

portunity to signiﬁcantly improve the accuracy and

relevance of job recommendations (Le Barbanchon

et al., 2023). Even though existing commercial tools

that offer such recommendations exist (e.g. sonara.ai,

dream.jobs), they lack the element of interactivity

with the user and the integration of external resources

relevant to the user skills, which are useful for spe-

https://orcid.org/0000-0003-3742-7123

https://orcid.org/0000-0003-0658-5065

ciﬁc disciplines, such as the development activity of

the user for software engineering positions.

In this work, we aim to create a chatbot-based web

platform that efﬁciently matches job seekers with rel-

evant opportunities based on a detailed analysis of

their skills (hard and soft skills) and preferences. The

chatbot aims to gather user information in a conver-

sational manner, making the process of adding prefer-

ences and skills more engaging. The main contribu-

tion of the work lies in enhancing the efﬁciency of the

job search and recruitment process by demonstrating

how AI-driven chatbots can be utilized to personal-

ize job seeking. By analyzing a large number of job

listing from online sites, as well as developers’ CVs

and GitHub activities, the platform offers at the cur-

rent state of implementation personalized job recom-

mendations to software engineers, thereby increasing

the chances of successful employment. Nevertheless,

more disciplines will be added in the future. The job

listings are collected from ergodotisi.com for Cyprus

and glassdor.com for United States. The matching of

CV and job listings is based on the user’s previous

years of experience, soft skills, hard skills, education,

and other personal preferences, such as location and

company. We have evaluated the chatbot via a small-

scale user study focusing on assessing its usability

and effectiveness in matching jobs and resumes, and

Fotiadis, P., Kapitsaki, G. M. and Papoutsoglou, M.

A Job Finder Chatbot-Based Web Platform: A Use Case for Software Engineers.

DOI: 10.5220/0013749000003985

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 21st International Conference on Web Information Systems and Technologies (WEBIST 2025), pages 197-204

ISBN: 978-989-758-772-6; ISSN: 2184-3252

197

the overall user satisfaction (Chowdhary and Chowd-

hary, 2020).

Chatbot Source Code Availability. The job

ﬁnder chatbot is available for replication purposes on-

line in a GitHub repository: https://github.com/

CS-UCY-SEIT-lab/Job-Finder-Chatbot.

The remainder of the text is structured as follows.

Section 2 presents relevant related work, while Sec-

tion 3 describes the design choices and the implemen-

tation tools of the chatbot, including use demonstra-

tion. The user evaluation is presented in section 4,

while section 5 discusses the main ﬁndings. Finally,

section 6 concludes the work.

2 RELATED WORK

Chatbots. Chatbots function effectively across var-

ious ﬁelds, from customer support to personal assis-

tance, making them a crucial asset in numerous ap-

plications (Nagarhalli et al., 2020). In customer ser-

vice, they can offer 24/7 assistance, handling inquiries

and resolving issues promptly (e.g. chatbot.com).

In e-commerce, chatbots enhance the shopping ex-

perience by offering personalized recommendations

and support (e.g. maisieai.com). In healthcare, they

can triage symptoms and provide health-related in-

formation (Shinde et al., 2021), whereas they have

also been used for career counseling purposes (Suresh

et al., 2021). In software engineering, they have been

used to assist in understanding and using Open Source

Software licenses (Shittas et al., 2025).

Skills, Job Matching and Recommendations.

Prior works have focused on extracting skills from

job advertisements, using both automatic and man-

ual techniques. A pre-trained LinearSVC model was

used for identifying skills in advertisements in Ger-

man (Grüger and Schneider, 2019), while manual

analysis was used for ﬁnding soft skills in job ad-

vertisements for software positions in Cyprus (Kapit-

saki et al., 2024). Earlier research has also focused

on gathering software developers’ prior experience

from different sites in one location, in order to assist

employers (Greene and Fischer, 2016; Kapitsaki and

Foutros, 2017).

In job matching, an algorithm considering

seeker’s preferences has been introduced by Zhou

et al. (Zhou et al., 2019). Zero-Shot Learning and

speciﬁc pretrained models (e.g. all-MiniLM-L12-

v2) were used in order to align job descriptions with

candidate proﬁles (Kurek et al., 2024). Job adver-

tisements from Avenga were employed for this pur-

pose, while the used dataset contains also previous

mappings manually created by recruiters. When it

Table 1: Comparison of current work with online job seek-

ing systems.

Features sonara.ai simplify.jobs careerﬂow.ai dream.jobs Our work

Chatbot

function-

ality

✓ ✓

Analyze

re-

sume/CV

✓ ✓ ✓ ✓

Large

variety of

jobs in

database

✓ ✓

Considers

hard skills

✓ ✓ ✓ ✓ ✓

Considers

soft skills

✓ ✓

Use of

GitHub

✓

comes to personalized job recommendations, using

the largest online job board in Sweden, it was found

that recommending a vacancy to a job seeker in-

creases the seeker’s possibility to work at that speciﬁc

place (Le Barbanchon et al., 2023).

Concerning online tools, the recent escalation in

the development of AI made it possible to create tools

that help, personalize, and boost the process of ﬁnd-

ing the best-ﬁtting jobs for the seekers. Sonara,

Sim-

plify,

Careerﬂow

and Dream.jobs

are main such

tools that offer different subscription plans. Sonara

and Careerﬂow work by analyzing user’s CV for

years of experience and hard skills and based on that

information ﬁlters the job listings that are posted on

the side of the employer. Simplify uses a chatbot

to collect information such as hard skills, location,

salary, and preferred company. This system has a lot

of companies uploading their job listings directly to

them. Dream.jobs also has a small number of job list-

ings, but the system’s CV analyzer takes into consid-

eration hard skills and soft skills. Table 1 provides a

comparison of the features of the above systems and

the current work.

3 DESIGN AND

IMPLEMENTATION

3.1 Platform Overview

The implemented job ﬁnder chatbot-based web plat-

form aims to match the abilities of a software engi-

neer based on their CV and their GitHub activity with

available job advertisements in online sites. The plat-

https://www.sonara.ai/

https://simplify.jobs/

https://www.careerﬂow.ai/

https://dream.jobs/

WEBIST 2025 - 21st International Conference on Web Information Systems and Technologies

198

form initially analyzes users’ CV and GitHub proﬁle

in order to extract hard skills and soft skills. GitHub

was chosen, as it summarizes well developer’s prior

experience in OSS projects and has been used also in

prior works for this purpose, functioning as an up-to-

date CV for developers (Greene and Fischer, 2016;

Kapitsaki and Foutros, 2017). Then, with the use of

the chatbot, the users provide information on the de-

sired job location, company, type of employment (full

or part time), their years of experience and any ad-

ditional hard skills and soft skills they want to add.

Finally, using the collected information the platform

ﬁlters the available job listings that are collected and

analyzed from glassdoor.com and ergodotisi.com, in

order to provide to users best matches.

3.2 Sources of Job Listings

The selection of glassdoor.com and ergodotisi.com as

primary sources for the data collection was inﬂuenced

by their structural similarities and extensive job list-

ings. Both platforms present job advertisements in a

concise list format, with each listing linked to a sep-

arate URL detailing the full job description. Glass-

door.com, being one of the leading job advertisement

websites in the United States, and ergodotisi.com,

holding a similar standing in a European country

(Cyprus), were good candidates for compiling a com-

prehensive and inclusive dataset of job posts. This

way two geographical areas were added in the plat-

form, whereas additional sources for other locations

can be added. Data are collected at regular intervals

to ensure the platform considers recent job posts.

Before settling for those two sources, other plat-

forms, such as LinkedIn and Monster were considered

and examined. However, they were both excluded for

the following reasons. LinkedIn features a sophisti-

cated search engine and complex website structure,

but extracting information is not allowed. Monster

was found to have a limited number of listings for

software engineering positions that were targeted for

the initial implementation of the chatbot-based web

platform.

3.3 Data Collection and Extraction

Concerning the data collection process, Selenium

was used to automate the browsing process to collect

job listing data from the online platforms. Beautiful-

Soup4

was used alongside Selenium to extract spe-

ciﬁc content from web pages after they were loaded

https://www.selenium.dev/

https://pypi.org/project/beautifulsoup4/

and interacted with by Selenium, taking into consid-

eration the job listing structure from the sites. Light-

cast

offers a taxonomy in hard and soft skills con-

sisting of more than 33,000 skills accessible via the

Lightcast’s Open Skills API. This service was pre-

ferred over creating a skills taxonomy from scratch or

employing other Machine Learning or Natural Lan-

guage Processing (NLP) techniques to extract skills,

as Lightcast updates these data constantly, ensuring

that any new hard skill or technology that emerges

is added in the database. The Lightcast API is then

used to return the relevant soft and hard skills found

within each job post description (using as input the

job post text), so that these can be subsequently

matched against the user’s skills. The ﬁnal structure

of the data collected from job posts are stored in JSON

(JavaScript Object Notation) format so that they can

subsequently be matched against the developer proﬁle

and preferences (note however, that job postings from

ergodotisi.com are used only for matching purposes

and are not further stored).

From the side of the developer, data are collected:

1) from the user (via the interaction with the chatbot),

2) the uploaded CV and 3) the user’s GitHub pro-

ﬁle. Uploading a CV generally allows a job match-

ing system to analyze comprehensive details about

a user’s educational background, skills, and accom-

plishments, and is the traditional approach used also

in the online systems (e.g. sonara). Integrating

GitHub data enables the system to assess in addi-

tion the user’s coding projects, contributions to pub-

lic repositories, and technical skills, which are often

not fully captured in a traditional CV. The system can

thus, identify not only the programming languages

that the user knows but also their recent activity and

expertise level in using those languages in real-world

projects. The information collected from the user via

the chatbot are speciﬁcally: location(s) of desired job

positions, job type, preferred company, years of ex-

perience, education information, hard skills, and soft

skills. These data are also stored in a JSON schema

as job listings data.

The data retrieved and used for the job listings

and the user are all listed in Table 2. Concerning the

user, ﬁelds beginning with info are provided by the

user (via the chatbot), while the rest are retrieved from

the developer’s CV and GitHub proﬁle (programming

languages captured in hard skills and user’s location

are extracted from GitHub, while hard skills and soft

skills are retrieved from the user’s CV). For the em-

ployment type, we are using the following options: 1)

full or part time, 2) hybrid, remote or ﬂexible, 3) per-

manent or temporary.

https://lightcast.io/

A Job Finder Chatbot-Based Web Platform: A Use Case for Software Engineers

199

Table 2: Data used for the job listing and the user.

Name Data type

Job listings data (J)

URL string

location string

company name string

employment type string array

years of experience string

hard skills string array

soft skills string array

User data (U)

ID string

username string

info_location string array

info_job type string array

info_company string array

info_years of experience string array

info_education level string array

info_education type string array

info_hard skills string array

info_soft skills string array

cv_hard skills string array

cv_soft skills string array

github_hard skills string array

github_location string

3.4 Skills Matching

When all data from the job adverts and the user are

available, skill matching is performed using the two

sets of data: one set comprises of the hard or soft

skills of the user (U) and the second set contains the

respective hard or soft skills from the job adverts (J).

We decided to perform a match on skills on keywords

level, as the extended taxonomy of Lightcast covers a

very large number of skills. For this purpose, we are

relying on Jaccard similarity:

J(J, U) =

| J ∩U |

| J ∪U |

(1)

The two sets are considered relevant for the user,

and hence the respective job advert, if there is a value

of 0.2 or higher for either the hard skills or the soft

skills matching. We have opted for this threshold af-

ter experimentation with the matching process in the

sources of job adverts, but the threshold can be up-

dated (e.g. if more sources are added).

3.5 Implementation Tools and Chatbot

Flow

The Rasa

framework was used at the core of the

implementation as it was adopted as the chatbot im-

plementation framework. It was chosen due to its

popularity and since it has an Open Source edi-

tion. Rasa uses a number of steps for understanding

user’s intent and performing the appropriate action.

It combines tokenization, feature extraction, and so-

phisticated classiﬁcation and decision-making mech-

anisms, to empower the application to handle user in-

teractions effectively, recognizing intents and extract-

ing entities, with the above features:

1. User input: The job ﬁnder chatbot collects de-

tailed user inputs across several categories criti-

cal to job matching, including geographic loca-

tion, job type preference (e.g. full time), years

of experience, educational background, soft skills

and hard skills, as detailed in subsection 3.3. It

engages users in a comprehensive dialogue, en-

suring that all relevant information is gathered to

support personalized job recommendations.

2. NLP analysis: The job ﬁnder chatbot successfully

interprets and processes user responses, such as

interpreting "Alabama" as a location and recog-

nizing "5 years" as the level of experience. The

NLP component of Rasa has been tuned in the

framework of the current work with relevant con-

ﬁguration ﬁles offered to understand and extract

meaningful information from user inputs.

3. User feedback mechanism: Each stage of the in-

teraction with the chatbot allows for user conﬁr-

mation or correction, which is crucial for main-

taining the accuracy of user data.

4. Data integration: The job ﬁnder chatbot-based

web platform is an integrated platform that takes

user data (e.g. skills and experience) from various

sources (chatbot, CV, GitHub) and maps them to

the available job postings (then shows to the user

links to job listings tailored to the user’s proﬁle).

The backend of the web platform is leveraging the

collected data to fetch relevant job opportunities

from external job sites.

For the chatbot functionality, Rasa relies on a

number of key conﬁguration ﬁles, including do-

main.yml, nlu.yml, and stories.yml. Understanding

user’s intent is vital for the above process, and specif-

ically items 2 and 3. The nlu.yml ﬁle contains lookup

tables with expected words that the user can give and

variations of answers that the user might give using

https://rasa.com/

WEBIST 2025 - 21st International Conference on Web Information Systems and Technologies

200

the lookup tables’ words. This ﬁle is used to cre-

ate the natural language model which the chatbot is

based on, recognizing the user’s answers and decid-

ing on the chatbot’s story ﬂow. A subset of the possi-

ble user intents when collecting soft skills is presented

below, with the text in square brackets ([]) referring

to the parameter from the user provided text and the

text in parenthesis (()) referring to the entity’s name,

i.e. (soft_skills) in the example intent:

− i n t e n t : c o l l e c t _ s o f t _ s k i l l s

e x am pl e s : |

− I am good a t [ teamwork ] ( s o f t _ s k i l l s )

− [ l e a d e r s h i p ] ( s o f t _ s k i l l s )

− [ f a s t l e a r n e r ] ( s o f t _ s k i l l s ) and

[ l e a d e r s h i p ] ( s o f t _ s k i l l s )

− [ f a s t l e a r n e r ] ( s o f t _ s k i l l s ) , [ l e a d e r s h i p ]

( s o f t _ s k i l l s ) and [ teamwork ] ( s o f t _ s k i l l s )

− I e x c e l i n [ com mu n ic at i on ] ( s o f t _ s k i l l s )

and [ t e a m b u i l d i n g ] ( s o f t _ s k i l l s ) .

− My s t r e n g t h s i n c l u d e [ c r i t i c a l t h i n k i n g ]

( s o f t _ s k i l l s ) and [ pr obl em s o l v i n g ] ( s o f t _

s k i l l s ) .

− I p o s s e s s s t r o n g [ a n a l y t i c a l s k i l l s ]

( s o f t _ s k i l l s ) and [ i n n o v a t i o n ] ( s o f t _ s k i l l s )

c a p a b i l i t i e s .

− [ S e l f − mo t i v a t i o n ] ( s o f t _ s k i l l s ) i s one o f

my c o r e q u a l i t i e s .

− I have e x p e r i e n c e w it h [ v i r t u a l t ea m s ]

( s o f t _ s k i l l s ) and [ kn o wled ge t r a n s f e r ]

( s o f t _ s k i l l s ) .

− [ C o n s u l t i n g ] ( s o f t _ s k i l l s ) ,

[ r e p o r t w r i t i n g ] ( s o f t _ s k i l l s ) , and

[ b u d ge t i ng ] ( s o f t _ s k i l l s ) a r e my e x p e r t i s e

a r e a s .

− I am known f o r my [ r e l i a b i l i t y ]

( s o f t _ s k i l l s ) and [ a b i l i t y t o mee t

d e a d l i n e s ] ( s o f t _ s k i l l s ) .

− P r o f i c i e n t i n [ q u a l i t y a s s u r a n c e ]

( s o f t _ s k i l l s ) and [ p r o g r e s s r e p o r t i n g ]

( s o f t _ s k i l l s ) .

[ . . . ]

The domain.yml ﬁle contains intents and entities

names that can be extracted from user’s input. In our

case, the entities correspond to the user data that can

be collected from the chatbot as presented in Table 2

(starting with info). Stories.yml ﬁle deﬁnes a diagram

of the chat ﬂow and how the chatbot must proceed af-

ter each question is answered and what information it

should expect from the user to provide. We are rely-

ing on a main story ﬂow, which contains a large num-

ber of actions and intents, in order to collect all re-

quired user data sequentially. All conﬁguration ﬁles

are available on the GitHub repository of the platform.

For the main implementation of the platform,

MongoDB was also used for data storage, Docker was

used for making the chatbot deployment easier, and

Flask was used for the server-side logic and database

interactions. We also employed JSON Web Tokens

(JWT) that are deﬁned by the open standard RFC

7519 for securely transferring information, while we

relied on React JS for frontend development pur-

poses.The interconnection between the main tools is

depicted in Figure 1.

Figure 1: Interaction of implementation tools.

3.6 Use Demonstration

Basic interactions with the chatbot-based web plat-

form are depicted in Figure 2 and Figure 3. When the

user is logged in the system, the available page layout

allows the user to refresh the current chat (by clicking

“Refresh Chat”), create a new chat (by clicking “New

Chat”), upload his/her CV (using “Resume/CV De-

tails”), trigger the GitHub analysis (by giving his/her

GitHub username) and navigate to old chats. An ex-

ample of the respective analysis results after the user

has uploaded his/her CV and has triggered the GitHub

analysis is shown in Figure 2. When the user adds

his/her GitHub username, the backend uses GitHub’s

API to detect all programming languages that are in

the repositories the user has contributed to, and to also

expert the user’s location if this information is avail-

able in the user proﬁle on GitHub. The text of the CV

is also analyzed to detect hard and soft skills (rely-

ing as aforementioned on the taxonomy provided by

Lightcast). Those data are used along with the data

collected through the chatbot conversation in order to

ﬁnd the most ﬁtting job adverts for the user.

Concerning the chatbot, user’s temporary data that

are stored during a chatbot use are deleted every time

the user logs in or clicks on “Refresh Chat” option.

An example interaction ﬂow with a user is depicted

in Figure 3. The user is interested in job posts in Al-

abama, whereas Larnaca is detected as user’s location

from GitHub, so both locations are used for the job

listings recommendations. The user is interested in

full type employment, has 5 years of experience and

is not interested in speciﬁc companies. The user then

provides information on his/her education level, hard

and soft skills, and at the end the chatbot recommends

a speciﬁc job post to the user.

4 USER EVALUATION

Survey Creation. The chatbot-based web platform

for job recommendations has been evaluated with the

A Job Finder Chatbot-Based Web Platform: A Use Case for Software Engineers

201

Figure 2: CV analysis in job ﬁnder chatbot.

Figure 3: Interaction example with the job ﬁnder chatbot.

participation of users from the University of Cyprus.

The user evaluation aimed to gather feedback on the

usability, effectiveness, and overall user experience

with the web platform. The questionnaire consisted

of 15 questions that cover the main parts presented

in Table 3, including prior experience with job search

and the feedback from the interaction with the imple-

mented platform. Questions used are mainly multi-

ple choice, 5-Likert case and open ended questions.

Potential participants within the Computer Science

Department of University of Cyprus were contacted

via email communication. Participants were informed

that no personal data would be collected in the survey,

that their replies would be used solely for research

purposes and would not trace back to the participant

in any way. In order to answer the questionnaire, par-

WEBIST 2025 - 21st International Conference on Web Information Systems and Technologies

202

Table 3: Main parts of user evaluation questionnaire.

Section # questions Question examples

Background, job search behavior and preferences 4 What is your current employment status? (multiple choice)

Previous experience with job ﬁnder tools 4 In which industry/ﬁeld are you seeking employment? (open)

How many days per week do you dedicate to looking for job opportunities? (open)

Willingness to use AI-driven job ﬁnder tools 2 Would you use an AI chatbot that ﬁnds you personalized job listings based on the details

given on the chatbot, GitHub and CV that you provide? (multiple choice)

Feedback on chatbot usage 5 JobFinder chatbot was easy to use. (5-Likert scale)

(after interacting with the chatbot) I trust the job positions recommended to me by the chatbot. (5-Likert scale)

Figure 4: Users’ replies to main questions of evaluation.

ticipants gave their consent to the above.

Results. 21 users participated in the evaluation

in total. 13 (61.9%) of them were employed, while

3 (14.3%) were seeking employment and 5 (23.8%)

were students. Data concerning the hours users spent

on job search were collected (second section of the

questionnaire), with an average value of 5.4 hours per

week indicated by the users (standard deviation: 3.56

hours) and with some participants spending even 10

hours on this task weekly.

Concerning the experience with using the chatbot-

based web platform, most participants found it easy

to use: 13 (61.9%) participants agreed strongly and 8

(38.1%) agreed. Replies’ distribution to some of the

main questions on the chatbot interaction are shown

in Figure 4. The chatbot was generally well-received,

particularly in terms of usability, trustworthiness in

job recommendations and the length of the job search

process. An important question was whether users

would prefer this chatbot-based approach over the tra-

ditional job search and most of the users indicated a

preference for the chatbot: 15 (71.4%) participants

agreed or agreed strongly. Adding more features,

such as automatic job application by the chatbot on

behalf of the user, were also considered useful. Such

features were among the participants suggestions for

improving the chatbot.

5 DISCUSSION

We have gathered overall positive feedback from the

users, which allows also to transfer the current imple-

mentation to other disciplines beyond software engi-

neering. This would require considering potentially

also other sources of information for skills acquisi-

tion besides GitHub that is tailored to software de-

velopment. The main matching performed based on

the chatbot interaction with the user does not require

any adaptations for this transfer. Concerning the data

sources used, the accuracy of the recommendations

provided by the chatbot depends heavily on the com-

prehensiveness and reliability of the information it re-

ceives from the users. However, we argue that it is

vital to use more than one sources of information. If

the chatbot relies solely on conversational data pro-

vided during interactions with the user and does not

have access to additional detailed sources like CVs,

the matching might not be as accurate.

Limitations. In terms of data collection for job

adverts, we are limited by the terms of service and the

structure of respective sites, and for this reason it is

not feasible to use any site for job adverts collection.

Concerning the software development activity, we are

relying on GitHub to gather relevant data but this may

not reﬂect the user’s activity on all skills level, as it

refers mainly to programming language skills. More-

A Job Finder Chatbot-Based Web Platform: A Use Case for Software Engineers

203

over, the evaluation has been limited to the population

of University of Cyprus, so main results might differ

if a wider population of software engineers employed

in the industry had tested the chatbot.

6 CONCLUSIONS

In this work, we have presented a job ﬁnder chatbot-

based web platform, that assists users to get personal-

ized recommendations for their software development

job search, considering their preference and skills,

and using different data sources (CV and GitHub pro-

ﬁle). The chatbot can be used as a starting point

for the development of similar interactive job seek-

ing systems, and its initial small-scale user evaluation

is promising in this respect. Future work will expand

the chatbot’s algorithms and databases to support a

broader range of job categories beyond software en-

gineering, while new sources of job adverts covering

more countries will be added. Moreover, participants

from the software industry will also be recruited for

evaluation purposes. An extended evaluation will al-

low to draw more conclusions on its comparison with

traditional job searching techniques.

REFERENCES

Aluas, M., Angelis, L., Arapakis, I., Arvanitou, E., Geor-

giou, K., Gogos, A., Jahn, M., Kehagias, D., Kordoni,

V., Macaluso, S., et al. (2024). Skillab: Skills matter.

In 2024 50th Euromicro Conference on Software En-

gineering and Advanced Applications (SEAA), pages

491–498. IEEE.

Chowdhary, K. and Chowdhary, K. (2020). Natural lan-

guage processing. Fundamentals of artiﬁcial intelli-

gence, pages 603–649.

Greene, G. J. and Fischer, B. (2016). Cvexplorer: Identi-

fying candidate developers by mining and exploring

their open source contributions. In Proceedings of

the 31st IEEE/ACM international conference on au-

tomated software engineering, pages 804–809.

Grüger, J. and Schneider, G. J. (2019). Automated analysis

of job requirements for computer scientists in online

job advertisements. In WEBIST, pages 226–233.

Kapitsaki, G., Chatzivasili, L., Papoutsoglou, M., and

Galster, M. (2024). An exploratory study on soft

skills present in software positions in cyprus: a

quasi-replication study. In Proceedings of the 18th

ACM/IEEE International Symposium on Empirical

Software Engineering and Measurement, pages 200–

211.

Kapitsaki, G. M. and Foutros, P. (2017). Dear developers,

your expertise in one place. In 2017 43rd Euromicro

Conference on Software Engineering and Advanced

Applications (SEAA), pages 371–374. IEEE.

Kurek, J., Latkowski, T., Bukowski, M.,

Swiderski, B.,

Ł˛epicki, M., Baranik, G., Nowak, B., Zakowicz, R.,

and Dobrakowski, Ł. (2024). Zero-shot recommenda-

tion ai models for efﬁcient job–candidate matching in

recruitment process. Applied Sciences, 14(6):2601.

Le Barbanchon, T., Hensvik, L., and Rathelot, R. (2023).

How can ai improve search and matching? evidence

from 59 million personalized job recommendations.

Evidence from, 59.

Nagarhalli, T. P., Vaze, V., and Rana, N. (2020). A re-

view of current trends in the development of chat-

bot systems. In 2020 6th International conference

on advanced computing and communication systems

(ICACCS), pages 706–710. IEEE.

Shinde, N. V., Akhade, A., Bagad, P., Bhavsar, H., Wagh,

S., and Kamble, A. (2021). Healthcare chatbot system

using artiﬁcial intelligence. In 2021 5th International

Conference on Trends in Electronics and Informatics

(ICOEI), pages 1–8. IEEE.

Shittas, G., Kapitsaki, G. M., and Papoutsoglou, M. (2025).

Licy: A chatbot assistant to better understand and se-

lect open source software licenses. In Mannion, M.,

Männistö, T., and Maciaszek, L. A., editors, Proceed-

ings of the 20th International Conference on Evalu-

ation of Novel Approaches to Software Engineering,

ENASE 2025, Porto, Portugal, April 4-6, 2025, pages

573–581. SCITEPRESS.

Suresh, N., Mukabe, N., Hashiyana, V., Limbo, A., and

Hauwanga, A. (2021). Career counseling chatbot on

facebook messenger using ai. In Proceedings of the

International Conference on Data Science, Machine

Learning and Artiﬁcial Intelligence, pages 65–73.

Zhou, Q., Liao, F., Chen, C., and Ge, L. (2019). Job rec-

ommendation algorithm for graduates based on per-

sonalized preference. CCF Transactions on Pervasive

Computing and Interaction, 1:260–274.

WEBIST 2025 - 21st International Conference on Web Information Systems and Technologies

204