Automated Code Review Assistant

Sadish Sendil Murugaraj, Rama Krishna Vardhan Taddi, Jagan Allu and Pallavi Rapolu

Department of Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and

Technology, Avadi, Tamil Nadu, India

Keywords: Automated Code, Artificial Intelligence, Data Privacy.

Abstract: Automated code review is playing an important role in modern software development and the combination of

artificial intelligence is making it more and more important. This project employs artificial intelligence and

machine learning to scan for bugs, vulnerabilities and adherence to ethical norms. It will deliver automated

feedback on things like the algorithmic underpinning, availability to minorities and data privacy, evaluating

whether the software harmonizes with the tenets of equity and inclusion. The designed system will be multi-

language to use in the different programming environment. It will also be fully embedded within the CI/CD

pipelines to offer feedback as early as possible during the development cycle. By integrating a human rights-

based methodology into the code review process, the Tool helps to ensure that software products are also

compliant with ethical standards, thus, striving for transparency and accountability in technology

development. Closing the gap between social justice and technology, the importance of this project lies in

building sustainable and inclusive digital building blocks.

1 INTRODUCTION

The release of an Automated Code Review Assistant

for human-rights-based projects is a monumental step

towards combining ethics with technical excellence

for the future of our software development.

Automated code review tools employ machine

learning and artificial intelligence to scan the source

code for errors, weaknesses, and any deviations from

established criteria. The same goes with Human

Rights where such tools can be customized to

contribute to the alignment of software systems with

core ethical principles, protection of sensitive data,

and even promotion of equitable participation and

inclusion. These tools can save time and minimize

human error by ensuring best practices through

automated code review.

Human rights work often includes handling

sensitive information, so security and privacy are of

utmost importance. As an example, an automated

code review assistant can do static and dynamic code

analysis to find vulnerabilities like those which could

lead to data integrity issues or exposing individuals to

harm. In other words, it can review algorithms for

potential biases or discriminatory practices, allowing

the software to meet ethical standards. For example,

the tool can apply accessibility guidelines to make

sure that marginalized communities are included, in

keeping with global human rights standards.

Furthermore, the assistant may seamlessly

integrate into the current improvement workflows, in

addition to into CI/CD pipelines, giving actual time

feedback on a number of degrees throughout the

coding course of. This allows builders to address

problems before the improvement of lifecycle,

reducing costs and improving significant software

quality. The device can also generate specific

opinions that highlight ethical concerns, safety

hazards and practical recommendations for

remediation. The assistant integrates one's talents into

the upgrade route, promoting transparency and

responsibility in generation creation.

In conclusion, the automatic code study assistant

that we have designed for human rights obligations is

based on technical acting with ethical training. It

guarantees that no longer is the shape of software

program structures satisfactory assembly beneficial

necessities but additionally maintaining thoughts of

equity, safety, and inclusivity. Combining technology

and social justice, it empowers builders to construct

solutions that make an optimistic contribution to

society and adhere to international human rights

criteria.

Murugaraj, S. S., Taddi, R. K. V., Allu, J. and Rapolu, P.

Automated Code Review Assistant.

DOI: 10.5220/0013880000004919

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

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

195-201

ISBN: 978-989-758-777-1

195

2 RELATED WORKS

There has been a big deal of interest in automated

code evaluation assistants in latest-day years,

indicating a want for greater green software program

great guarantee practices. This phase of essays

applied works that learn the improvement,

implementation, and impact of these out in world

contexts.

As a result, automated code evaluation has

developed into a vital software for software program

application development to assist the demanding and

tedious technique of analyzing the supply code for

high quality and integrity. The other team adopts a

machine learning and tool watching ammend of AI

in order to research code, find thickening issues, and

provides feedback to builders. The automation of

advantageous elements of code assessment allows

builders to be targeted on extra complicated duties,

improving conventional overall performance and

productivity.

Recently, there has been work done on more

advanced techniques to automatically evaluate code

in a variety of ways. For example, researchers have

used software dependency graphs as decoration for

the reading of code shape records. It corresponds to

translating code into graph sequences the usage of

algorithms like PDG2Seq that permits fashions to

understand code superb judgment greater efficiently.

Models like these can set up statements in natural

language and advise code adjustments enhancing

development performance via giving builders greater

intuitive remarks. We also explored the application

of deep reading (DL) techniques with the goal of

automating finer-grained code quality assessment

tasks, decreasing dependence on guide reviews and

improving consistency.

In order to automate code, evaluate procedures in

recent years, machines learning strategies have

become increasingly used. Machine studying

methods for example, can streamline and improve

critiques by deciphering software program structures.

Additionally, massive language fashions (LLMs)

have proven encouraging effects in industry

functions, aiding in bug localization and enriching

code fine awareness. These models can research

explicitly and generate higher-pleasant opinions

makes them treasured gear in software program

improvement Moreover Candle et al. This tar portrait

paves the walk for lender improvements, should be

people-posed, focused, and balanced, to objurgate

entrenched impediments on directed approaches of

automating metamorphic tribute duties in

improvement, automatic code test measures were

added precisely into software program improvement

workflows to decorate overall performance, and

nice. For instance, the CORE (Code Review Tool)

tool automates examine suggestions for code

modifications to iterate better performance networks

primarily based totally on modifications code and

their respective critiques. But while automatic

equipment offers a few advantages, they're now not

an alternative to human judgment. Research

highlight the value of combining automated tools

with human intervention to ensure quality code.

As computerized code assessment maintains to

conform, future research should recognition on

improving the accuracy and reliability of those

equipment. This might also involve integrating extra

state-of-the-art tool studying models and leveraging

large datasets to enhance model performance shows

in Table 1. Additionally, information a way to

efficaciously stability automated gear with human

oversight may be crucial for maximizing the

blessings of automatic code examine. By advancing

in those regions, computerized code evaluation can

play an increasing number of crucial roles in

enhancing software program application development

techniques, making sure better-nice code and faster

development cycles.

3 METHODOLOGY

The proposed approach for growing an Automated

Code Review Assistant tailor-made to human rights-

targeted software program application software

program initiatives con- sists of a multi-section

technique to make certain ethical compliance,

accessibility, and safety. The tool will combine static

and dynamic code evaluation techniques to come

upon vulnerabilities, insects, and moral problems

inside the supply code. Machines getting to know

algorithms can be employed to find out types of

functionality bias, privateness violations, or

discriminatory practices embedded in the code.

Additionally, predefined rule gadgets aligned with

international human rights necessities will guide the

automatic study method.

The assistant can be included into CI/CD

pipelines to provide actual-time feedback

subsequently of development, making sure non-stop

monitoring and development. It will help multi-

language programming environments and use natural

language processing (NLP) strategies to analyze

comments and documentation for inclusiveness and

readability. To beautify usability, the assistant will

generate specific critiques highlighting problems,

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

196

their human rights implications, and actionable hints

for remediation.

This approach emphasizes collaboration through

way of permitting developers to engage with the tool

via customer- remarkable interfaces in Integrated

Development Environments (IDEs). Furthermore, it

will embody feedback loops to refine its normal

regular overall performance based on consumer input

and evolving human rights guidelines. By embedding

one’s abilities, the mission targets to create a robust

tool that not nice complements code nice but

moreover guarantees that software program

development aligns with moral and human rights

standards.

1. Data Collection

2. Study Design

3. Static Code Analysis

4. Natural Language Processing (NLP) Models

A. Automated code review data collection

B. Study Design

The proposed approach for developing an automatic

code overview assistant leverages a scientific method

to integrate synthetic intelligence (AI) into the code

study machine. This approach is designed to

beautify code pleasant, reduce take a look at time,

and help builders in maintaining excessive coding

necessities. The following outlines the vital element

components of the proposed technique:

Table 1: Dataset Description.

Categor

Data Point Description

Code Quality

Metrics

Code Analysis Depth

Measures how thoroughly the static analysis code is being

used.

Number of Defects

Foun

Tracks the number of issues reported including bugs, logic

errors, style violations.

Chan

e Failure Rate Percenta

e of code chan

es leadin

to errors

ost-mer

Code Duplication Identifies redundant or duplicate code blocks.

Review Efficienc

Review Time Time taken from submission to feedback or a

roval.

Pull Request Size Analyzes size of pull request in lines of code.

Throu

ut Number of reviews com

leted over a

eriod.

Reviewer Loa

Number of open pull requests assigned per reviewer.

Workflow

Integration

Integration with

CI/CD

Compatibility with CI/CD pipeline.

Version Control

System

Compatibility with tools like GitHub, GitLab, and

Bitbucket.

Customization Customizable Rules

Ability to define and enforce team-specific coding

standards.

1. System Integration: Development of IDE

Extension: Create an extension for well-known

Integrated Development Environments (IDEs) which

includes Visual Studio Code (VS Code). This

extension will permit developers to provoke code

reviews right away interior their coding surroundings.

2. Code Snippet Capture: Implement functionality to

seize code snippets or complete files that want

examine, sending them to a backend server for

processing.

3. Backend Processing: Server Architecture: Set up a

backend server the usage of cloud offerings (e.G.,

AWS EC2), so as to contend with requests from the

IDE extension.

API Communication: Establish verbal exchange

protocols among the IDE extension and the backend

server the usage of RESTful APIs. This lets in for

green records transfer and processing. (e.G., Chat-

GPT, Gemini API) for reading the captured code. The

models will check out the code for functionality

insects, style improvements, and safety

vulnerabilities.

4. AI Model Analysis: Utilization of Generative

AI Models: Leverage modern generative AI

models.

5. Feedback Generation: The AI model will generate

action- able remarks based totally at the assessment,

alongside tips for exquisite practices and vital

corrections.

Feedback Display: User Interface Design: Design an

intuitive customer interface in the IDE extension

that offers comments in a clean and actionable

format.

Interactive Feedback Mechanism: Enable builders to

interact with the comments, permitting them to take

Automated Code Review Assistant

197

delivery of or adjust recommended adjustments right

away inside their coding surroundings.

6. Continuous Learning and Improvement: Data Col-

lection for Model Training: Collect records from

consumer interactions and comments on the AI-

generated pointers to continuously decorate the

version’s accuracy and relevance.

Integration with Historical Data: Incorporate ancient

code evaluate facts to beautify the AI’s functionality

to provide context-aware suggestions tailor-made to

specific coding styles and assignment necessities.

7. Evaluation and Testing: Pilot Testing with Users:

Conduct pilot trying out with real customers in each

open-supply project and business settings to assess

the effectiveness of the device. Performance Metrics:

Measure overall performance primarily based on

metrics along with reduction in review time, accuracy

of hints, user delight, and improvement in code

exceptional.

Figure 1: Overview of Proposed Methodology.

C. Static Code Analysis

Static code evaluation is a pivotal issue of automatic

code assessment, supplying a systematic technique to

analyzing source code without execution. This

technique includes the usage of computerized gadgets

to scrutinize the codebase for functionality issues,

which consist of coding requirements violations,

safety vulnerabilities, and logical mistakes. By

integrating static code evaluation into the

improvement workflow, groups can pick out and deal

with problems early inside the improvement cycle,

thereby reducing the time and fee associated with

debugging and safety. This approach is specifically

useful in ensuring code best, safety, and

maintainability, as it enables placed into effect

particular requirements and practices, discover

vulnerabilities which include SQL injection and

buffer overflows, and beautify average software

reliability.

Moreover, static code evaluation aligns with the

“shift left” locating out motion, which emphasizes

integrating finding out and evaluation tools earlier

inside the software program improvement lifecycle to

trap problems in advance than they come to be full-

size issues. By leveraging static code assessment,

developers can build higher-extremely good

software applications, restrict technical debt, and

beautify improvement performance, in the long run

contributing to faster and further reliable software

program releases.

D. Detection Using langchaing

Figure 2: Lang Chain Architecture.

Figure 1 shows the proposed technique for

computerized code assessment entails an entire

approach to enhancing code exceptional and standard

universal performance through the strategic use of

automated machines and software programs. This

approach starts off evolved with the mixture of static

code assessment system, which check supply code for

compliance with predefined hints and notable

practices, figuring out issues together with coding big

violations, protection vulnerabilities, and ordinary

common overall performance bottlenecks. These

systems are carried out via scanning the codebase to

come across capability issues, supplying builders with

actionable remarks to rectify identified troubles right

away. The lang chain architecture is shown in Figure

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

198

A key difficulty of this tool is the use of automated

code assessment suites, which provide more superior

abilities than conventional linters. Tools like

SonarQube and Code Climate assist more than one

programming language and can be covered into non-

prevent integration/continuous deployment (CI/CD)

pipelines, imparting close to actual-time feedback to

builders. This integration permits the early detection

of bugs and vulnerabilities, ensuring that code is

dependable and solid in advance than it reaches

manufacturing.

To maximize the blessings of automatic code

evaluation, the proposed technique emphasizes the

importance of blending automated machine with

human oversight. While computerized tools excel at

identifying habitual problems, human reviewers are

vital for addressing complex accurate judgment

mistakes and ensuring that code aligns with venture-

unique necessities and commercial company right

judgment. By balancing automation with human

judgment, improvement organizations can leverage

the strengths of each strategy to decorate code

notably, reduce manual assessment efforts, and

accelerate the improvement cycle.

Furthermore, this device advocates for the

adoption of Agile improvement methodologies, in

which automatic code examine performs a important

characteristic in facilitating faster and extra green

iterations. By automating code reviews, businesses

can align with Agile necessities of handing over

exceptional software application software program in

shorter cycles, promoting collaboration and remarks

amongst corporation humans. This approach not

satisfactory improves code first-class however

moreover fosters a lifestyle of non- prevent

improvement internal development companies.

Overall, the proposed methodology for automatic

code evaluation gives a based framework for

enhancing software improvement processes with the

aid of leveraging generation to streamline code

evaluation, improve high-quality, and decrease

development time. By integrating automatic gear into

the development workflow and balancing them with

human oversight, teams can make certain that their

software meets the very best requirements of first-

class and reliability.

The textit F1 Score is the harmonic imply of

Precision and Recall, supplying a unmarried

measure that balances each metric, mainly useful

while the class distribution is imbalanced. The dataset

used for automated code evaluation is a entire series

of code snippets and related metadata, designed to

beautify the education and overall performance of AI

models in identifying code troubles. This dataset is

compiled from several properties, which encompass

network repositories, GitHub public repositories, and

StackOverflow, ensuring a big variety of coding

patterns and actual-global examples are represented.

The information consists of examples of correct code,

compatibility troubles, commonplace average

performance problems, and runtime errors, each

meticulously classified to facilitate powerful version

schooling. Advanced preprocessing strategies which

encompass records augmentation and normalization

are completed to decorate dataset amazing and

variety. For instance, facts augmentation involves

strategies like flipping code snippets and injecting

minor mistakes to simulate real-global coding

situations, even as normalization ensures consistency

in code formats across the dataset.

This approach is not the simplest aid in

schooling. AI models however also afford valuable

insights for builders throughout the overview method,

permitting them to address capability troubles greater

correctly. Additionally, datasets like those supplied

through Tufano et al. Consist of pre- submission code,

reviewer comments, and post-overview code

adjustments, supplying a rich context for expertise the

code evaluation procedure and enhancing version

accuracy.

Finally, the text Mean Average Precision (MAP)

is the average of the Average Precision (AP) across

all classes. The AP for each class is calculated as the

vicinity underneath the precision-do not forget curve,

which is an indispensable of the precision over the

consider values.

4 EXPERIMENTAL RESULTS

AND DISCUSSION

In us examine on automated code review, we

determined numerous key findings that highlight the

effectiveness and boundaries of these tools.

Automated code evaluation tools were found to

noticeably enhance the rate and performance of the

evaluation method through figuring out easy issues

which include code fashion inconsistencies and trivial

errors.

For example: tools like Code grip and Sonar Cube

proven high accuracy in detecting code smells and

safety vulnerabilities, presenting developers with

actionable remarks to improve code first-class.

However, regardless of these advantages, depending

totally on automatic gear can cause an illusion of

great manipulation, as they will omit complicated

Automated Code Review Assistant

199

logic mistakes or issues associated with code

rationale and person necessities.

Our experiments additionally discovered that deep

learning- based totally based models can be powerful

in automating unique code evaluating tasks. For

example, Deep Review, a version of the usage of deep

multi-example studying, confirmed promising effects

in figuring out capacity issues in open-supply

projects, with upgrades in both F1 score and AUC

metrics. Additionally, massive language fashions

(LLMs) have been used to beautify code review

efficiency by providing natural language feedback

and guidelines for code changes. However, these

models aren’t without barriers, as they are able to

generate false positives and negatives, which require

human intervention to resolve.

The integration of automatic code evaluation

equipment into continuous integration/non-forestall

deployment (CI/CD) pipelines changed into found to

offer close-to-time feedback, allowing builders to

cope with issues right away. This method no longer

best speeds up the improvement cycle but

additionally fosters a culture of excellent guarantee

in the development crew. Furthermore, combining

computerized code evaluations with practices like

pair or mob programming can in addition decorate

code excellently thru ensuring that human oversight

enhances automated checks.

In phrases of practical applications, our

observation discovered that automated code

evaluation equipment can substantially lessen the

time and effort required for manual critiques,

permitting developers to awareness on greater

complicated duties. However, it’s miles essential to

stability automated gear with human judgment to

ensure complete extremely good manipulation.

While automatic gear excels at detecting positive

sorts of troubles, they cann ot replace the nuanced

understanding and information that human reviewers

convey to the method.

Overall, our experimental outcomes underscore

the functionality of automatic code examine

equipment to enhance software improvement

techniques. By leveraging that equipment efficiently

and integrating it with human oversight, developers

can enhance code first-rate, lessen development time,

and decorate popular software reliability. Future

research needs to reputation on refining the ones

system to restriction faux positives and negatives and

to better address complex code problems that require

human understanding. Figure 3 and 4 shows the

comparison with an efficient code review platform.

Figure 3: Comparison with Efficient Code Review Platform

-1.

Figure 4: Comparison with Efficient Code Review Platform

– 2.

5 CONCLUSIONS

In quit, pc-coded code evaluation is a trans-formative

technique in software application software

development, providing several blessings that

embellish each daily performance and best. Through

the use of automatic de-vice, developers are able to

quickly turn into cognizant concerning capacity

issues along with syntax errors, bugs, and protection

weaknesses, to ensure that code meets installed

requirements and astonishing practices. This

automation no longer just saves time with the

valuable resources of manner of lessening the

dependence on guide assessments but additionally

encourages means of lifestyles of uniformity and

cooperation inner betterment corporations. In spite of

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

200

stressful circumstances along with pretend positives

and negatives, incorporating auto- mated code check

tools along traditional guide reviews offer a solid

platform for strengthening desirable code superb. As

software program responsibilities continue to evolve

in complexity and size, embracing computerized code

review methods could be crucial to passing on superb

applications effectively while reducing dangers

related to security loopholes and program- ming bugs.

Finally, the strategic application of computerized

code analysis tool can appreciably speed up the

improvement process, enabling teams to attain

awareness on more complex and advanced additives

of coding, thus improving the depend- ability and

performance of software programs with utility

software systems.

6 FUTURE WORK

Future work in the realm of automated code review is

poised Future paintings in the realm of automatic

code assessment is poised to be trans formative,

pushed with the useful resource of upgrades in

artificial intelligence and massive language fashions

(LLM). One promising area of studies includes the

improvement of self-analyzing LLM models that

might constantly improve via the usage of using

studying comments from builders, adapting their

pointers to better align with institution opportunities

and undertaking-specific requirements.

Further enhancements could probably incorporate

improving accuracy through schooling AI models on

historical code evaluation facts to enhance the AI’s

capability to offer context conscious guidelines tailor-

made to unique coding patterns and mission

requirements. Also, permitting for personalization of

coding requirements to align AI-driven pointers with

mission- unique fantastic practices is probably useful.

REFERENCES

A. Bacchelli and C. BIRD, ‘’Expectations, outcomes, and

challenges of modern code review,” in 2013 35th

International Conference on Software Engineering

(ICSE). IEEE, 2013, pp. 712-721.

https://doi.org/10.1109/ICSE.2013.6606617

A. JHindle, E.T. Barr, M. Gabel, Z. Su, ane P. Devanbu,

“On the naturalness of software,’’ Communications of

the ACM, Vol. 59, No.5, 2016, pp. 122-131.

https://earlbarr.com/publications/naturalness_cacm.pdf

A. Vaswani, N. Shazeer, N. parmar, J.Uszkoreit, L. Jones

et al., “Attention is all you need,” arXiv: 1706.03762,

2017. https://doi.org/10.48550/arXiv.1706.03762

A. M. Alshahrani et al., “Automating Code Review”, IEEE

Conference Publication, 2024https://doi.org/10.1109/I

CSECompanion58688.2023.00053

A. M. Alshahrani et al., “CORE: Automating Review

Recommendation for Code Changes,” IEEE, 2024,

https://doi.org/10.1109/SANER48275.2020.9054794

Czerwonka, M. Greiler, and J. Tilford, ‘’Code reviews do

not find bugs. How the current code review best

practice slows us down,” in IEEE/ACM 37th

International Conference on Software Engineering,

Vol. 2. IEEE, 2015 pp. 27-28 https://dl.acm.org/doi/

10.5555/2819009.2819015

D.S. Mendonca and M. Kalinowski, “An empirical

investigation on the challenges of creating custom static

analysis rules for defect localization,” Software Quality

Journal, 2022, pp. 1-28. https://doi.org/10.37190/e-

Inf250102

F. Huq, M. Hasan, M.M.A. Haque, S. Mahbub, A. Iqbal

et al., “Review4Repair: Code review aided automatic

program repairing’’, Information and Software

Technology, Vol. 143, 2022, p. 106765.

https://doi.org/10.48550/arXiv.2010.01544

J. Smith et al., “Towards Automating Code Review

Activities,” arXiv,2021. https://arxiv.org/pdf/2021.02

518.pdf

M. Staron, W. Meding, O. Soder, and M. Back,

“Measurement and impact factors of speed of reviews

and integration in continuous software engineering,’’

Foundations of Computing and Decision Sciences, Vol.

43 No. 4, 2018, pp. 281-303.http://dx.doi.org/10.1515

/fcd s-2018-0015

M. Staron, M. Ochodek, W. Meding, and O. Soder, “Using

machine learning to identify code fragments for manua

review’’, in 46th Euromicro Conference on Software

Engineering and Advanced Applications (SEAA),

IEEE, 2020, pp. 513-516 http://dx.doi.org/10.1109/S

EAA51224.2020.00085

M. Hasan, A. Iqbal, M.R.U. Islam, A. Rahman, and A.

Bosu, “Using a balanced scorecard to identify

opportunities to improve code review effectiveness: An

industrial experience report’’, Empirical Software

Engineering, Vol. 26, No. 6, 2021, pp. 1-34.

https://doi.org/10.48550/arXiv.2101.10585

N. Fatima, S. Nazir, and S. CHUPRAT, “Knowledge

sharing, a key sustainable practice is on risk: An insight

from modern code review,” in IEEE 6th International

Conference on Engineering Technologies and Applied

Sciences (ICETAS). IEEE, 2019, PP. 1-6.

http://dx.doi.org/10.14569/IJACSA.2020.0110160

Umut Cihan, “Automated Code Review in Practice,” arXiv,

Vol. 2412.18531 2024. https://doi.org/10.48550/arXiv

.2412.18531

Umut Cihan, “Automated Code Review in Practice,” Vol.

2412.18531, 2024. https://doi.org/10.48550/arXiv.241

2.18531

Ying Yin, et al, “Automatic Code Review by Learning the

Structure Information of Code Graph’’, Big Data

Analytics and Intelligent Computation to Advance

Novel Applications, 23(5), 2551. https://doi.org/10.33

90/s23052551

Automated Code Review Assistant

201