PES UNIVERSITY

DEPARTMENT OF Electrical & Electronics Engineering

100 Feet Ring Road, BSK 3rd Stage, Bangalore-560085

InternshIp report

Submitted by:

PRAVEEN KUMAR
PES1UG22EE804

8th Semester B.Tech

(Electrical and Electronics)

Under the guidance of

Suraj Srivastava, Accure AI
 PES UNIVERSITY
DEPARTMENT OF Electrical & Electronics Engineering
100 Feet Ring Road, BSK 3rd Stage, Bangalore-560085

Certified that the internship work entitled “Machine Learning Intern” is a Bonafide work
carried out by PRAVEEN KUMAR bearing SRN PES1UG22EE804 in partial fulfillment
for the award of degree of Bachelor of Technology in Electrical and Electronic by PES
University, at Accure AI, Bangalore during the period Feb-May in the academic year 2024-
2025. The internship has been approved as it satisfies the academic requirements in respect of
internship work prescribed for the award of the degree of Bachelor of Technology.

Dr. Rex Joseph

Dean of Faculty
PES University
RR Campus, Bangalore

1
 DECLARATION

I PRAVEEN KUMAR Student of Electrical and Electronics Engineering Department from

PES UNIVERSITY , Bangalore, hereby I declare that the Internship report is submitted in
partial fulfillment of academic requirements for the award of getting certificate from the
respectable organization ACCURE AI. It has been carried out independently by me under the
guidance of SURAJ SRIVATSAVA(Accure AI).

PLACE: Bengaluru PRAVEEN KUMAR

DATE: 28/05/2025 (PES1UG22EE804)
Electrical and Electronics
Engineering

2
 Acknowledgement

First and foremost, I am deeply thankful to Mr. Mogishuddin Raza, Chief Operating Officer
at Accure AI, for providing me the opportunity to be part of such an innovative and forward-
thinking organization. His leadership and vision for applied AI deeply inspired me throughout
my internship.
I extend our heartfelt gratitude to late, Dr. M.R. Doreswamy, the visionary Founder and
Chancellor of PES University, for his unwavering guidance and support.
I also express sincere appreciation to Prof. Jawahar Doreswamy, Pro-Chancellor, and Dr. J.
Surya Prasad, Vice Chancellor, for entrusting us with the privilege of being a part of this
prestigious institution.
My profound thanks go to Dr. K.S. Sridhar, Registrar of PES University, for his invaluable
support in creating an environment conducive to innovation and learning.

I am also especially grateful to Mr. Suraj Srivastava, my mentor at Accure AI, for his
invaluable guidance and support during the course of the internship. His mentorship was
instrumental in helping me understand the intricacies of machine learning workflows and
apply them effectively using the company’s proprietary platforms. His encouragement and
insights helped me navigate both the learning and project phases of the internship with
confidence.
I acknowledge with deep respect the encouragement and insightful guidance of chairperson
Dr. Rex Joseph, Chairperson of the Department of Electrical & Electronics Enginering.
I would also like to acknowledge and thank Pushpa K R, Assistant Professor, Department of
Electrical and Electronics Engineering, PES University, for her academic supervision and
encouragement throughout the internship process.
Finally, I am thankful to all faculty members of the Department of Electrical and Electronics
Engineering, PES University, for equipping me with the foundational knowledge and skills
that enabled me to make the most of this internship opportunity.
This experience would not have been possible without the support of all the above individuals,
and I remain deeply appreciative of the opportunities I received during my time at Accure AI.

3
Contents

1 Introduction
1.1 Objective
1.2 Scope of Report
2 Company Profile
3 Role and Responsibilities
3.1 Collaboration
4 Technical Contributions
4.1 Week 1-5: Certification Course- No Code End-to-End Data Engineering and ML
4.2 Week 6: Generating Contract Project with SecureGPT
4.3 Week 7-12: Project Using Accure Momentum
5 Machine Failure Detection
6 Finalised Outcomes and Conclusion
6.1Finalised Outcomes
6.2Conclusion
7 Certificate of Internship

4
`
1. Introduction
During the winter internship term of 2025, I had the chance to intern at Accure AI, an
exciting AI startup based in Bengaluru that focuses on generative AI, enterprise automation,
and applied machine learning. This experience gave me hands-on exposure to how machine
learning is applied in real-world scenarios, especially in emerging areas like procurement and
document intelligence.
Throughout the internship, I worked with Accure AI’s in-house platforms, Secure GPT and
Accure Momentum, to develop data-driven solutions. As an electrical and electronics
engineering student with a growing passion for analytics and data science, this role helped me
connect my technical background with the world of AI. I gained a clearer understanding of
practical ML workflows, sharpened my coding and data handling skills, and had the
opportunity to contribute to real AI product development. It was a valuable step in exploring
how engineering and AI can come together to solve real problems.

1.1 Objective
The main objective of this internship was to gain hands-on experience in building and
evaluating machine learning models while contributing to actual use-case implementations
for clients. My role involved using Secure GPT to perform Contarct Generation and Accure
Momentum to develop classification models on various datasets. In addition, I was exposed
to natural language processing (NLP) tasks and the preprocessing of legal and procurement
documents. I also learned about the practical implications of data privacy and the importance
of secure, explainable AI solutions in enterprise environments.

1.2 Scope of the Report

This report presents a detailed account of the activities and learnings during my 3-month
internship at Accure AI. It outlines the tools and software platforms used, the machine
learning concepts explored, and the contributions made toward model development and
evaluation. The report begins with an overview of Accure AI’s work in AI/ML technologies,
followed by my weekly tasks, key projects, learnings in ML workflows, and the final
outcomes. The report also reflects on the skills developed and how the experience shaped my
understanding of AI's impact in industry.

5
 2. Company Profile
Accure AI is a next-generation AI startup headquartered in Bengaluru, India. The company is
focused on making AI explainable, secure, and operational for businesses across industries.
Accure AI develops enterprise-grade AI solutions that enable companies to harness the power
of data and automation with transparency and accountability. It is backed by a team of
experienced technologists and data scientists and operates with a strong emphasis on
innovation, ethical AI, and domain-specific adaptability.
Accure AI's product portfolio includes:
 Secure GPT: A document intelligence and NLP platform that applies generative AI to
securely analyse and compare large volumes of legal, financial, or procurement
documents.
 Accure Momentum: A no-code ML platform that allows users to build, test, and
deploy machine learning models for classification, regression, and clustering tasks
with ease. It is built for rapid model development, performance tracking, and
scalability.
The company works with clients in sectors such as legal tech, manufacturing, logistics, and
finance, offering tailored AI solutions for automation, anomaly detection, forecasting, and
intelligent document processing. What differentiates Accure AI is its commitment to data
privacy and explainable AI (XAI), ensuring that AI-driven decisions can be audited and
interpreted by business users and stakeholders.
Accure AI’s mission is to make AI trustworthy and usable at scale, focusing on building
responsible AI systems that align with business goals while maintaining compliance and
transparency. Its collaborative and innovation-driven work culture has helped it become a
sought-after startup for AI internships and talent looking to enter the data science field.

6
 3. Role and Responsibilities
At Accure AI, I was designated as a Machine Learning Intern, with a focus on exploring the
applications of generative AI and automated model development tools. The internship was
structured to provide me exposure to Accure AI's suite of enterprise software, including
Secure GPT and Accure Momentum. My tasks included data preprocessing, document-based
Contract Generation using AI, model training and evaluation, and testing platform
functionalities in real-world scenarios.
My responsibilities evolved over the course of the internship and involved active
participation in the entire model development pipeline. I worked closely under the guidance
of my assigned mentor at Accure AI and collaborated with software engineers and data
scientists to understand how artificial intelligence models are developed, evaluated, and
deployed in a secure enterprise setting.
I was also tasked with preparing documentation and feedback reports regarding model
performance, and learned how to translate domain-specific requirements—such as
procurement or legal standards—into structured data inputs for machine learning models. In
the process, I improved my knowledge of classification algorithms, accuracy metrics, and
how AI can be fine-tuned for business applications.

3.1 Collaboration:
Throughout my internship, Accure AI fostered a highly collaborative environment. Weekly
sync-up sessions allowed me to share progress, clarify doubts, and receive constructive
feedback. These interactions helped me gain clarity on concepts such as prompt engineering,
model interpretability, and the ethical handling of data within AI systems. While most of my
work was independently driven, the team provided constant support and encouraged me to
ask questions, experiment with models, and take ownership of mini-projects. The software
teams were approachable and explained how features within Accure’s tools are tailored for
different use cases across industries. Working alongside a multi-disciplinary team gave me
insight into how machine learning integrates into broader software ecosystems. It also helped
me appreciate the balance between technical development and usability, especially in
enterprise-facing platforms where explainability, interface design, and security are just as
important as algorithm performance. Overall, my role during this internship was not limited
to mere observation; I actively contributed to real use cases involving document intelligence
and data modelling. The experience strengthened my interest in data science and gave me a
strong foundation to further explore machine learning technologies in both academic and
professional settings.

7
 4. Technical Contributions
During my 12-week internship at Accure AI, I had the opportunity to engage in a structured
learning and project-based experience that enabled me to apply machine learning concepts
using Accure’s proprietary platforms. The internship began with an intensive onboarding
phase through a certified course in data engineering and machine learning, followed by
hands-on projects using Accure’s Secure GPT and Momentum tools. Each phase of the
internship contributed to building a strong foundation in data processing, model development,
and AI-powered document intelligence.

4.1 Week 1-5: Certification Course – No Code End-to-End Data Engineering

and ML
The initial five weeks of the internship were dedicated to completing a structured online
course titled “No Code End-to-End Data Engineering and Machine Learning”, offered
through Accure AI’s internal learning platform. This course served as an essential foundation
for interns before they began working on client-based or internal projects. It was designed to
help us understand the fundamental components of the data lifecycle, including data sourcing,
cleaning, transformation, model training, evaluation, and deployment—all within a no-code
environment.
The course content was organized into modules, each consisting of pre-recorded video
lectures, reading materials, interactive assignments, and multiple quizzes. Topics covered
included:
 Introduction to data pipelines and the ETL (Extract, Transform, Load) process
 Basics of supervised and unsupervised machine learning
 Understanding feature engineering and data transformation techniques
 Model selection and evaluation using accuracy, precision, recall, and F1-score
 Introduction to no-code platforms for ML model creation and automation
 Ethics, fairness, and explainability in AI
To complete the course and receive certification, it was mandatory to score above a minimum
threshold across all module assessments and submit final project tasks simulating a simplified
ML workflow. The quizzes and assignments were designed to test conceptual clarity and the
ability to apply AI tools without traditional coding. This approach was instrumental in
enabling interns from diverse technical backgrounds to contribute meaningfully to later
projects.
Successfully completing this course not only marked the completion of the onboarding phase
but also equipped me with the technical readiness required to use Accure’s tools in real-world
applications. By the end of Week 5, I had received my course completion certificate, having
developed a clear understanding of the ML lifecycle and its no-code implementation.

8
 Certificate of Completion

4.2 Week 6: Generating a contract Project using Secure GPT

In Week 6, I transitioned to my first generative contract automation task using Secure GPT,
Accure AI’s document intelligence platform powered by generative AI. This project focused
on automatically generating procurement-related legal contracts by leveraging a structured
reference agreement and minimal user-created input.
The goal of the project was to demonstrate how Secure GPT could be used to generate
consistent, legally formatted contracts in real-world enterprise use.
Key tasks and learnings during this week included:
 Uploading and preprocessing procurement documents in various formats (PDF,
DOCX, etc.)
 Understanding how large language models can be applied securely within enterprise
environments
 User input parsing and mapping
 Prompt engineering: Crafting effective natural language prompts to extract
structured insights
 Generating complete contract document using Secure GPT
 Reporting output inconsistencies and model performance feedback to the product
team

9
This project helped me gain exposure to how privacy-preserving AI solutions are essential in
industries dealing with sensitive or confidential information. My contributions included
testing multiple document samples, suggesting refinements in prompt templates, and creating
a basic report outlining model accuracy and coverage on legal terms across different cases.

Secure GPT

Working with Documents on Secure GPT

10
4.3 Week 7–12: Project Using Accure Momentum (Machine Failure Detection)
In the final phase of my internship, I worked extensively with Accure Momentum, a no-code
machine learning platform, to build and evaluate a classification model aimed at Machine
failure detection based on Predictive Maintenance dataset. This project involved exploring
various stages of the ML pipeline—from data uploading, cleaning, and transformation to
model selection, training, and validation—all within a no-code environment.
My contributions included identifying relevant features, handling missing data, selecting
suitable classification algorithms, and tuning model parameters to improve performance
metrics such as accuracy and F1-score. I also documented the entire process, analysed model
outputs using built-in explainability tools, and shared performance reports with the team.

11
 5. Machine Failure Prediction
5.1 Introduction
the project is for process of developing end-to-end machine learning system of predicting machine
failure for preventive maintenance. We will explore how to use Momentum AI to perform various
steps from data acquisition to model training, deployment and monitoring.

5.2 Objective
The primary objective of this analysis is to:

train a machine learning model to predict if the machine fails given certain conditions.

This is AutomL classification problem.

5.3 Dataset
We will utilize the dataset available at the UCI Machine Learning Repository [1][2]. The following
information is copied from the UCI website.

5.4 Abstract
The AI4I 2020 Predictive Maintenance Dataset is a synthetic dataset that reflects real predictive
maintenance data encountered in industry.

5.5 Attribute Information

Below is a description of each attribute in the dataset:

Column Name Description

UID unique identifier ranging from 1 to 10000

Product ID consisting of a letter L, M, or H

Air temperature generated using a random walk process later normalized to a standard

12
 Column Name Description

deviation of 2 K around 300 K

Process generated using a random walk process normalized to a standard deviation of 1

temperature K, added to the air temperature plus 10 K.

Rotational speed
calculated from a power of 2860 W, overlaid with a normally distributed noise
[RPM]

torque values are normally distributed around 40 Nm with a f = 10 Nm and no

Torque [Nm]
negative values.

The quality variants H/M/L add 5/3/2 minutes of tool wear to the used tool in
Tool wear [min]
the process.

label that indicates whether the machine has failed in this particular datapoint
Machine failure
for any of the following failure modes are true.

TWF Tool wear failure

HDF Heat dissipation failure

PWF Power failure

OSF Overstrain failure

RNF Random failure

5.6 Data Wrangling and Exploration

Effective data analysis begins with preparing and understanding the dataset. This section outlines the
key steps involved in uploading, ingesting, and exploring the ai4i2020.csv file.

5.7 Uploading Data

The dataset, titled ai4i2020.csv, was uploaded to the Momentum platform by Accure AI. This CSV file
contains tabular data with comma-separated values and is ready for direct ingestion into the
platform's SQL or Python environment.

13
 Fig. Uploading Data

5.8 Checking File Upload

The file was successfully detected and displayed in the system, confirming a valid upload.

Fig. Uploaded CSV File

5.9 Ingesting Data

The uploaded dataset was ingested using standard CSV parsing procedures. Column headers were
automatically identified, and the data types were inferred correctly:

14
 Fig. Ingested Data

Fig. Log of Ingested Data

5.10 Exploring Dataset

After successful ingestion of the .csv file into Momentum, an initial exploration of the dataset was
conducted to understand its structure and content. Below are a few steps that illustrate
Momentum’s interactive and visual data exploration features using the ai4i2020.csv dataset.

15
 Fig. Data Exploration Result

5.11 Ad Hoc Analysis Using Interactive Query

Momentum supports Interactive Query, an ANSI-SQL compliant feature that allows you to run
custom queries on your data in real-time.

As with traditional RDBMS systems, the structure is hierarchical:

 The component name (e.g., "Ingester Output") acts as a database.

 The datasets are referenced using fully qualified names in the format:
in.<username>.<tablename>

For example, to count how many participants in the dataset have myopia, we use:

SQL QUERY1:

16
17
5.12 Creating Training and Test sets using transformer
There are many ways to create training and test sets from the dataset we ingested. This
section demonstrates how to use SQL compliant transformer to create the two sets for
machine learning training and test.
Assuming we want to create 80% training and 20% test sets from the original 10,000 records.
We also want to randomize the data so that the training and test sets are not biased. We will
create two transformers:
• The training set transformer will have a randomized 8000 records
• The test set transformer will have 2000 records that are not in the training set.

Transformation of data
Data Transformation serves as the foundation for generating actionable insights. Raw data is
often messy, inconsistent, or incomplete, which makes it unsuitable for decision-making
processes or training AI models. Additionally, data typically comes in various formats and
originates from multiple sources, such as spreadsheets, CSV files, databases, and APIs. Data
transformation is essential for cleaning, filtering, sampling, or converting this raw data into a
consistent and usable format. It also plays a key role in merging and integrating data from
diverse systems or sources. Within Accure Momentum, data transformation is powered by
SQL, offering both flexibility and precision to meet a wide range of data transformation
needs effectively.

18
Creating Training Set

Creating Test Set

19
5.13 Model Training and Testing
We utilize AutoML to automate the model selection and training process. AutoML
streamlines the end-to-end pipeline by automatically performing tasks such as data
preprocessing, feature selection, model selection, hyperparameter tuning, and evaluation.
This allows us to efficiently identify the best-performing classification model without manual
intervention. The process ensures consistent and optimized results across different datasets
and use cases, making it a powerful tool for scalable and accurate predictive modelling.

Training AutoML classification Model

20
21
22
Testing Model
A sample interactive query to analyze the prediction test result.

23
 6. Finalised Outcomes and Conclusion
6.1 Finalised Outcomes
Over the course of the 12-week internship at Accure AI, I had the opportunity to gain
significant exposure to real-world applications of artificial intelligence and machine learning
through both guided learning and independent project work. Successfully completing the
certified No Code End-to-End Data Engineering and Machine Learning Course marked
an important milestone, giving me a strong foundation in model development workflows,
data handling, and ethical AI principles. The certification not only validated my grasp of
theoretical concepts but also demonstrated my ability to apply them practically using no-code
platforms.
As part of my project using Secure GPT, I explored the potential of generative AI in
automating document comparison and analysis for enterprise applications. My primary
contribution involved conducting prompt-based comparisons of procurement agreements to
identify and highlight language-level differences between document versions. I analyzed the
outputs to generate structured insights and compiled detailed reports based on my findings.
These reports were shared with the internal product team and played a role in shaping product
feedback and guiding improvements to the software’s functionality for contract analysis and
document intelligence in enterprise settings.
In the second half of the internship, I worked on a classification model to predicting machine
failures cases using Accure Momentum, where I independently navigated the full lifecycle
of an ML project—from dataset cleaning and feature selection to model evaluation and
reporting. I applied classification techniques and experimented with hyperparameters to
arrive at a reliable and explainable model. This experience was pivotal in deepening my
confidence in handling applied ML tasks, even without coding, and understanding the role of
user-friendly tools in democratizing data science.
Overall, my work during this internship directly supported Accure AI's mission of building
secure, explainable, and accessible AI systems for real-world use cases.

6.2 Conclusion
My internship at Accure AI was a rewarding and transformative experience that strengthened
my technical and analytical capabilities while opening my perspective to the world of
artificial intelligence. The structured learning phase helped me build a solid conceptual base
in machine learning and data engineering, which I was able to apply meaningfully in two
major projects using Accure’s enterprise tools.
Through hands-on experience with Secure GPT, I learned about the practical deployment of
LLMs for document intelligence in industries like legal and procurement. Later, with the
Momentum project, I experienced how no-code tools can make advanced ML tasks
approachable, repeatable, and scalable—valuable skills for any aspiring data practitioner.

24
Beyond the technical learning, the internship also sharpened my ability to work
collaboratively, communicate findings clearly, and think critically about AI’s ethical
dimensions. The mentorship, structured guidance, and exposure to real-world challenges have
made a lasting impact on my career outlook. This experience has not only strengthened my
interest in the AI domain but also prepared me to contribute effectively to future data-driven
roles in industry or research.

25
7. CERTIFICATE OF INTERNSHIP COMPLETION

26