👤 About Me
Hello! I'm Pranav Agarwal, a pursuer of knowledge and innovation within the realm of technology. My journey began in the lively streets of India, where, as a young boy, I was fascinated by the magic of computers by spirited competitions in car racing games with my elder brother, sparking a passion for technology that continues to drive me forward.
At VIT University, my undergraduate studies in Computer Science allowed me to dive deeper into the mechanics of programming and data structures. It was here that I developed a keen interest in solving complex problems through code, leading to my first project— a car game built using C++.
My professional journey has been equally diverse, spanning several industries from game development, medicine to aerospace, each role enriching my expertise and broadening my perspective on how technology can be applied to different fields.
With a Master’s in Data Science at UC Irvine, I am on a quest to harness the power of machine learning to transform vast amounts of data into actionable insights. My goal is to innovate in ways that prioritize ethical considerations and privacy, ensuring that advancements in AI and machine learning are used responsibly.
Away from the keyboard, my passions are varied and vibrant. I find rhythm and poetry in hip-hop, the strategic depth and speed in Formula 1 racing, and the exhilarating competition of soccer. Each of these interests adds a layer of balance and joy to my life, fueling my work with renewed energy and inspiration.
🎓 Education
Bachelor of Technology in Computer Science
I earned my B.Tech in Computer Science and Engineering at Vellore Institute of Technology (VIT), India. My academic journey here was marked by engagement with core subjects like data structures, database management, and natural language processing. As an active member of ACM-VIT, I contributed to several projects, notably one focused on the automatic segregation of wet and dry waste using computer vision. This practical application of classroom learning fueled my interest in real-world technological solutions and prepared me for advanced studies.
Master of Data Science
With a Master of Data Science at the University of California, Irvine (UCI), where I achieved a GPA of 3.9. At UCI, I specialized in machine learning, artificial intelligence, and deep learning. My projects ranged from theoretical research to practical implementations, including an in-depth evaluation of foundational models for pathology in survival prediction. This program has honed my analytical skills to tackle complex data-driven challenges in the tech industry.
🔧 Skills & Certifications
Throughout my academic and professional journey, I have cultivated a diverse and robust skill set that spans various aspects of technology and data science:
- Programming Languages: Proficient in Python, C++, R, and SQL.
- Frameworks and Libraries: Experienced with PyTorch, TensorFlow, spaCy, Keras, Seaborn, and Pandas.
- Web Development: Skilled in using HTML, CSS, JavaScript, and Django for web applications.
- Cloud Technologies: Competent in managing and architecting solutions on AWS, including the use of EC2, S3, and RDS.
- Containerization and Orchestration: Knowledgeable in Docker and Kubernetes for developing scalable applications.
- NLP: Capable of building and deploying machine learning models, leveraging libraries like Hugging Face and LangChain.
- Database Management: Experienced in using SQL databases as well as NoSQL databases like MongoDB and Neo4J.
- Big Data Tools: Familiar with Apache Spark for large-scale data processing.
- Data Visualization: Proficient in creating insightful visualizations with Tableau and Python libraries like Matplotlib.
My certifications further bolster my technical capabilities:
- AWS Solutions Architect Associate: Certified in designing distributed systems on Amazon Web Services.
View Certificate - Applied Machine Learning: Certified by the University of Michigan on Coursera, focusing on practical machine learning implementation.
View Certificate
🔗 Open Source Contributions
Mozilla
I am an active contributor to Mozilla, focusing on projects that enhance web technologies and user privacy. My contributions help improve the performance and security features of Mozilla products, supporting a free and open web.
FOSSASIA
My work with FOSSASIA includes contributing to software that promotes the use of open source technology in Asia. These contributions span across various projects aimed at enhancing accessibility, education, and open source adoption.
OpnTec
At OpnTec, I contribute to the development of open source tools and technologies, enhancing innovation and collaboration within the tech community. My efforts focus on improving software solutions that facilitate open communication and technology sharing among developers and engineers worldwide.
🛠 Experience
Data Scientist | CVS Health
June 2025 - Present
I have just started my work here. The team is super supportive and helpful. I hope to achieve a lot at CVS!
Data Scientist | ImpactEdge, Inc.
Feb 2025 - Jun 2025
At ImpactEdge, I led the automation of sustainability document ingestion and parsing pipelines, reducing manual effort by 80% through the use of Selenium, LlamaParse, and Python. Extracted content was structured into a searchable Neo4j knowledge graph to enhance accessibility and retrieval.
I also improved hybrid information retrieval accuracy by 60% by integrating FAISS vector search, BM25 ranking, and graph traversal algorithms, deployed as a FastAPI-based microservice and containerized via Docker Compose.
Key Skills: Information Retrieval, Knowledge Graphs, Neo4j, FAISS, BM25, FastAPI, Docker Compose, Python, Selenium, LlamaParse.
Machine Learning Researcher | UCI AI Center
June 2024 - June 2025
At UCI's AI Center in Dr. Jana Lipkova’s OctoPath lab, I spearheaded the comprehensive evaluation of patch-level and whole-slide images utilizing diffusion models. My work, which utilized datasets from TCGA, CPTAC, and private cohorts, improved model accuracy to 89%.
I also increased accuracy for survival prediction targeting brain and lung cancers, delivering insights into the effectiveness of foundational models in clinical survival prediction.
.I applied PCA on the features extracted to reduce dimensionality down from 768 to the number of PCs that explain 90% variability. Brought base model accuracy to 81%
Key Skills: Machine Learning, GenAI, diffusion models, Data Analysis, Python, TensorFlow, Statistical Modeling.
Machine Learning Engineer Intern | Safran
July 2024 - Nov 2024
I saved $10,000 per aircraft per month by collaborating in the design and development of cloud-based data applications, specializing in machine learning models for predictive health management systems in aerospace equipment. This role involved intricate teamwork and technical skills to boost efficiency in developing solutions that predict equipment failures before they occur.
Further, I engineered and integrated LSTMs and ensembled models to enhance prototype systems, verifying their functional performance, and optimizing predictive algorithms within an IoT framework on AWS using SageMaker and Amazon ECR.
I also improved model accuracy to 96% by developing an ensemble model combining LSTMs for temporal data and XGBoost for categorical features to improve bias and variance.
Key Skills: Machine Learning, LSTMs AWS, SageMaker, Amazon ECR, IoT, Software Engineering, Data Analytics.
Data Scientist | Airbus
July 2021 - August 2023
As a Data Scientist at Airbus, I reduced monthly security alerts by 20% by employing analytical Bayesian methodologies using python to improve precision of alert systems by minimizing false positives and increasing precision.
I played a pivotal role in migrating 65% of departmental systems to AWS, enhancing operational efficiency and security.
My strategic implementations achieved annual cost savings of $70,000 by integrating a recommendation engine that optimized cloud resource allocation.
Additionally, I developed a comprehensive dashboard with Amazon QuickSight, providing actionable insights for strategic planning and project management.
Key Skills: Cloud Computing, AWS, Security Management, Data Analytics, Systems Engineering.
NLP developer Engineer Intern | Novartis
January 2021 - June 2021
During my internship at Novartis, I engineered an NLP-based chatbot using Azure Bot Framework that reduced response times by 1.7 hours, significantly improving customer service efficiency. It also replaced L1 support that cut cost by $1 million.
The implementation of a chatbot analyzer led to a 33% surge in usage by providing actionable insights for performance enhancement and a 23% increase in positive customer feedback through A/B testing of intents based on user feedback.
Key Skills: Natural Language Processing, Python, Machine Learning, Azure Bot Framework, Data Analysis.
📁 Projects
Lane Detection for Autonomous Driving Using Attention
This project implemented advanced lane detection techniques using PyTorch and OpenCV, crucial for improving the navigation systems of autonomous vehicles. The approach involved comparing a standard UNet model with a custom variant featuring residual blocks and attention mechanisms for enhanced lane marking detection accuracy that take into account the current data stream along with past knowledge to make predictions. The attention mechanism helps in the model focusing the relevant details of the image. The TUSimple dataset was used, showcasing multi-lane highway scenarios under varying conditions.
Key steps included preprocessing with grayscale conversion, Gaussian blur, and Canny edge detection, followed by random transformations to augment the dataset. The custom UNet model excelled at detecting lanes even in obstructed views, illustrating the advantages of attention-focused segmentation.
Technologies Used: Python, PyTorch, OpenCV, UNet, Attention Mechanisms
.gif)
View on GitHub | Read Article on Medium
Duplicate Detection in Job Postings using NLP and Milvus
The goal of this project was to solve the prevalent issue of duplicate job postings across multiple platforms. Using NLP techniques, I developed a system that could analyze and understand the semantic content of job descriptions. By implementing a vector search powered by Milvus, the system effectively identified and flagged duplicate postings, thereby enhancing the quality and trustworthiness of job boards.
Further development included optimizing the Sentence Transformers for better context capture and refining the cosine similarity calculations to improve the precision of duplicate detection. This refinement allowed the system to scale efficiently, adapting to new job postings and maintaining high accuracy over time. The project exemplifies how targeted AI solutions can address specific industry challenges, improving operational efficiency and user experience.
The application of cosine similarity also helped in the development of RAGs when finding the closes job description.
Key Skills: NLP, Milvus, Sentence Transformers, Python, Cosine Similarity, Vector Search.
Forest Fire Detection Using Classifiers and Transfer Learning
This project aimed to enhance forest fire detection capabilities through the use of sophisticated machine learning classifiers and transfer learning techniques. By integrating SVMs, KNN, and Naive Bayes with pre-trained models such as Inception and VGG19, I significantly improved the accuracy and speed of detection. The application of these advanced techniques allowed for early identification of fire outbreaks, potentially saving landscapes and lives by facilitating quicker response times.
The project also involved extensive data gathering and processing, where I implemented image recognition and processing technologies to identify fire characteristics from satellite imagery. The success of this project not only demonstrated the effectiveness of combining multiple machine learning strategies but also set a precedent for future environmental monitoring technologies.
Key Skills: Machine Learning, SVM, KNN, Naive Bayes, Transfer Learning, Image Processing, Python, Inception, VGG19.
Credit Reporting Trends Analysis
This project leveraged Python, Pandas, Seaborn, and NumPy to process and visualize data trends within credit reporting. Focusing on predictive analytics, it identified key trends affecting credit scores and market conditions, using comprehensive datasets from Equifax, Experian, and TransUnion.
By analyzing historical data, the project forecasted significant trends and their impacts on future credit activities, providing valuable insights for financial institutions and policymakers. This analysis was instrumental in developing strategies that align with evolving market dynamics.
Technologies Used: Python, Pandas, Seaborn, NumPy
Bombay Stock Exchange Equity Analysis
This project developed a comprehensive web application designed to monitor and analyze the daily performance of equities on the Bombay Stock Exchange. Utilizing real-time data scraping techniques, the system provided updated stock performance metrics that were crucial for day traders and long-term investors alike. The incorporation of machine learning algorithms allowed for predictive analysis, offering forecasts that helped users make informed investment decisions.
The backend of the application was powered by sophisticated data cleaning and analysis routines, ensuring data integrity and relevance. By employing advanced data visualization tools, the platform presented complex data in an accessible format, enabling users to quickly grasp market trends and variances. This project showcased the potential of integrating financial data with modern technology to create powerful analytical tools that empower users with actionable insights.
Key Skills: Data Scraping, Data Cleaning, Machine Learning, Python, JavaScript, Predictive Analytics, Data Visualization.
🌟 Extracurricular Activities
Winner of VITHACK'19
Awarded $1500 for the best project and $300 for the best website at VITHACK'19, a competitive hackathon focused on fintech innovations. My project featured a cutting-edge application that utilized computer vision to detect electronic devices in a home environment. By simply scanning the room with a smartphone camera, the system identified present electronics, enabling the fintech company to tailor personalized advertisements and product suggestions based on actual user possessions.
Winner of WTM GDG VIT'19
Developed a pivotal women’s safety application for WTM GDG VIT'19, which won accolades for its innovative approach to enhancing women's security in urban environments. The app collected data on past incidents of dangers faced by women and used this data to create a dynamic heatmap. It then provided intelligent routing suggestions to help women navigate through areas with the lowest risk, significantly boosting their safety in potentially hazardous locales.
Code for Good
Participated in Code for Good, an initiative dedicated to bringing basic computer science education to underprivileged schools. I engaged with students who lacked access to computers, teaching them fundamental computing skills and igniting an interest in technology that could potentially transform their futures.
Soccer
An enthusiastic participant in local soccer leagues, actively engaging in regional tournaments. Soccer not only enhances my physical fitness but also fosters a sense of teamwork and strategic thinking, skills that are invaluable both on and off the field.
