Yashashvini Rachamallu | Software Developer

Hi, I'm Yashashvini Rachamallu.

A

Yashashvini Rachamallu is a self-motivated and innovative computer scientist with a profound passion for programming and a keen aptitude for solving complex, real-world problems. Her expertise spans across machine learning, deep learning, computer vision, and natural language processing, reflecting her broad skill set and deep curiosity in exploring new challenges. She excels in initiating and propelling projects to successful completion, leveraging her diverse knowledge to create impactful solutions in AI.

About

Yashashvini Rachamallu is a driven and innovative software engineer with a passion for machine learning, large language models, and AI-driven solutions. As a Master of Science in Computer Science candidate at Michigan State University, she has maintained a perfect 4.0 GPA while contributing to impactful projects in areas such as ML optimization, data science, and e-commerce. Yashashvini excels at bridging theoretical knowledge with practical applications, delivering solutions that address complex challenges.

During her tenure at RICE F.W. Technologies, Yashashvini developed custom deep learning models and RAG-based AI chatbots, achieving a 60% reduction in document processing time and cutting operational costs by 50%. By leveraging Whisper for speech-to-text, LLaMA2 for advanced language modeling, and Azure OpenAI, she consistently delivered scalable, efficient solutions. Her work also involved large-scale data migration, cloud deployment, and implementing time-series forecasting models, resulting in a 20% increase in e-commerce sales.

In addition to her industry roles, Yashashvini has been actively engaged in research at Michigan State University. Her efforts led to a 40% improvement in crack detection accuracy for road infrastructure monitoring and accelerated geospatial data processing to over 1 million points per second. Her prior experience at Intel Corporation involved optimizing chip-placement workflows and reducing project costs by 70% through advanced ML-driven methods.

Yashashvini is passionate about continuous learning and problem-solving, always striving to push the boundaries of AI and data science. She seeks roles where she can apply her expertise in programming, machine learning, and cloud technologies to forward-thinking projects. Her ability to collaborate with teams, adapt to challenges, and deliver impactful solutions makes her a valuable asset in any innovative organization.

Experience

RICE F.W. Technologies Inc.

Machine Learning Engineer

KnowledgePlus – Digital Workspace

Decreased document processing time by 60% and operational costs by 50% by integrating Whisper for speech-to-text and LoRA-fine-tuned LLaMA2-70B for advanced document refinement.
Elevated efficiency via 4-bit quantization, facilitating cost-effective management of large-scale document workflows.
Developed a custom RAG-based AI chatbot using Azure OpenAI, Search AI Services, and Flask APIs for precise and rapid query resolution.
Implemented an Agile SDLC approach for chatbot development and deployment, cutting onboarding time and costs by 50%.
Automated document refinement workflows to significantly reduce manual intervention and operational delays.

WeCOMM – E-commerce Platform

Drove a 20% increase in annual sales by implementing a time-series forecasting model for inventory prediction, aligning stock with demand and preventing stockouts.
Migrated over 200,000 legacy MySQL records through a custom ETL pipeline, transforming data into a schema optimized for UI compatibility and production use.
Collaborated directly with clients to ensure migration efforts met business objectives, delivering both operational efficiency and strategic value.
Architected and deployed cloud infrastructure solutions including VM creation, SQL Server configuration, and KeyVault security setups for robust e-commerce operations.
Leveraged Azure ARM templates to automate server provisioning and infrastructure deployment, minimizing manual configuration efforts.
Performed comprehensive data preprocessing and attribute mapping, ensuring all legacy records were fully compatible with the updated system.
Executed secure cloud migrations, optimizing data access, reliability, and platform scalability.
Enabled seamless platform adoption by providing production-ready historic data, directly supporting real-time operational decision-making.

May 2024 – Jan 2025 | United States (On-site)

Michigan State University

Software Engineer

Software Engineer - ML Algorithms

Developed a geospatial classification system using Open3D, cKDTree, and multithreading to enhance processing efficiency.
Accelerated data analysis, achieving the ability to process over 1 million points per second, significantly improving performance.
Optimized geospatial workflows through advanced parallel processing techniques, enabling faster and more scalable solutions.

Graduate Teaching Assistant

Collaborated with professors and teaching assistants to improve lab exercises and create an internal course website for MSU, enhancing access to resources for 1,300+ students.
Managed course materials and coordinated with 90+ sections and teaching assistants, ensuring smooth course operations.
Streamlined workflows by improving communication between professors, TAs, and students, ensuring consistent delivery of instructional content across sections.
Leveraged automation to optimize exam preparation and course operations, improving efficiency and accuracy in managing large-scale course activities.

Software Engineer (Research Assistant)

Improved road crack detection by utilizing Mask R-CNN and Faster R-CNN, achieving highly precise segmentation and marking to streamline road maintenance operations.
Boosted maintenance efficiency with rapid output generation, enabling timely repairs and enhancing road safety.
Led a research initiative and collaborated with a team to optimize crack detection accuracy and processing speed.
Achieved 0.001-second output per image, enabling rapid-response solutions for quicker and safer road repairs.

Jan 2023 - Jan 2025

Intel Corporation

Graduate Technical Intern

Chip-Placement Optimization

Formulated the problem statement for optimizing chip-placement in ASIC workflows, focusing on leveraging graph-based machine learning techniques for enhanced accuracy and efficiency.
Developed a prototype integrating GraphSAGE neural networks, showcasing its superior ability to aggregate neighborhood information for precise chip component placement.
Built custom ETL pipelines using TCL scripting to extract and preprocess complex design data, enabling seamless integration into EDA tools and supporting model training.
Achieved a 30% improvement in wire length, power, and performance metrics, demonstrating the practical impact of GraphSAGE and ML-driven approaches in semiconductor design.

EDA Workflow Automation

Implemented efficient ML algorithms to analyze complex electronic design workflows, streamlining validation processes and reducing computational overhead.
Designed clustering algorithms for faster analysis and to ensure precise circuit simulation results.

Aug 2021 – Jun 2022 | Bengaluru, Karnataka, India

IIT Kanpur

AI Workshop by Ismiriti

Autonomous Bot Development

Developed a small-scale autonomous bot using NodeMCU, motors, and ultrasonic sensors, capable of detecting and avoiding obstacles with a 95% success rate.
Collected data by operating the bot via an app created using Blynk and MIT App Inventor, ensuring smooth control and real-time data transfer to a laptop.

Neural Network Deployment

Trained a neural network model on the obstacle detection data, achieving a 90% improvement in detection accuracy, and deployed it on a Raspberry Pi for autonomous operation.
This project demonstrated the integration of IoT technologies and machine learning for practical applications in robotics, with potential use in navigation, logistics, and automation tasks.

May 2019 – Jun 2019 | Kanpur, Uttar Pradesh, India

Projects

                
Covid-19 Classification 
                  An application helps in indentifiation of covid using MRI scans.
                
AccomplishmentsReached 97% accuracy in deep learning with CNNs and augmentation.
Used GANs to improve COVID-19 detection in lung scans.
Combined CNNs and GANs for higher COVID-19 identification accuracy.
Overcame data limits with augmentation and GANs, boosting performance.

Stock Market Prediction
                  An application helps in predicting the future rise or drop in stocks.
                
AccomplishmentsChose LSTM for top accuracy after comparing with GRU.
LSTM found best for accuracy in thorough forecasting method evaluation.
Developed a stock prediction system with 63% next-week accuracy.
 Boosted system with sentiment analysis, reaching 47% prediction accuracy.

Network Anomaly Detection
                  An analaysis on Network Anamoly Detection.
                
AccomplishmentsUsed GANs for resampling the Cup 99 dataset and applied PCA, LDA, and autoencoders for dimensionality reduction.
Achieved high accuracy in anomaly detection by evaluating different models.
Combined machine learning techniques for better cybersecurity defenses.
Analyzed the dataset deeply to improve anomaly detection strategies.

Solving Poisson equation
                  Optimized 2D Poisson equation solving with SOR, significantly cutting runtime via hybrid models.
                
AccomplishmentsDeveloped numerical methods for the 2D Poisson equation in science/engineering.
Used Successive Over Relaxation (SOR) in serial and hybrid models.
Highlighted hybrid (OPENMP+MPI) model's major runtime efficiency improvement.
Achieved significant computational efficiency optimization with the hybrid approach.

Language detection 
                  Developed a multilingual identification system with BERT, LSTM, achieving 86% accuracy.
                
AccomplishmentsBuilt language system with BERT, LSTM on 60 languages.
Analyzed four datasets, enhancing language recognition capabilities.
Integrated Bag of Words, Naive Bayes, Word Embedding techniques like Word2Vec for processing improvement.
Achieved up to 86% accuracy in language classification, identification.

Restaurant Recommendation System
                  Built recommendation system with 98.5% rating accuracy.
                
AccomplishmentsDeveloped Restaurant Recommendation system with help of Yelp dataset.
Applied regression, clustering for precise restaurant tips.
Achieved 98.5% accuracy in predicting business ratings.
Created effective restaurant recommendation system with high accuracy.

Image caption generation
                  An Caption Generator for Low-Light Images. 
                
AccomplishmentsDeployed two models, one for enhancement and other for captions generation.
Achieved robust captioning for images with difficult lighting conditions.
Achieved 0.6 - 0.7 BLEU scores for caption accuracy.
Model excels in linguistic accuracy, producing relevant captions.

Ulterior Website
                  The Ulterior, a website for reading books and some fun games.
                
AccomplishmentsCreated 'The Ulterior,' a literary site with free book access.
Used HTML, CSS, PHP, Bootstrap, and JavaScript for development.
Designed user-friendly and responsive interface for seamless reading.
 Connected readers globally, enhancing engagement and satisfaction.

Fake News Detection
                  Developed Fake News Detector, achieving 98% accuracy, tested real-world.
                
AccomplishmentsBuilt Fake News Detection with NLP and deep learning.
Achieved 98% accuracy on ISOT using LSTM and Word2Vec.
Tested in real-time with Times of India and Polifact content.
Reached 75% title accuracy, 66% for descriptions in real-world tests.

Similar Faces Detection
                  Developed similar facial recognition using Siamese Neural Networks.
                
AccomplishmentsBuilt facial recognition with Siamese networks on limited data.
Trained model using only the loss function effectively.
System identifies top three similar images for input.
Enables efficient facial similarity analysis through one-shot learning.

Trending YouTube video analysis 
                  Analyzed YouTube trends in six countries, revealing insights with 98% accuracy.
                
AccomplishmentsAnalyzed YouTube trends in six countries with Python, machine learning.
Used ensemble learning, data cleaning for video trend insights.
Achieved 98% accuracy predicting missing video categories.
Gained key insights from EDA and predictive models.

Hospital Management
                  Managed healthcare database with MYSQL, ensuring seamless patient record maintenance.
                
AccomplishmentsManaged healthcare database, maintained patient records for care planning.
Ensured accuracy across inpatient and outpatient services.
Implemented with MYSQL, tested with queries and triggers.
Enabled seamless data retrieval for hospital operations.

Yet Another Centralized Scheduler
                  Implemented YACS with dynamic allocation, achieving efficient big data scheduling.
                
AccomplishmentsImplemented YACS for efficient scheduling with dynamic task allocation.
Used Round Robin, Least Loaded, Random Selection algorithms.
Showcased task completion stats, mean and median values.
Demonstrated scheduling effectiveness for diverse big data jobs.

CPP Compiler
                  Built compact C++ compiler with yacc, lex, covering all phases.
                
AccomplishmentsBuilt compact C++ compiler with yacc and lex.
Covered all compiler design phases comprehensively.
Included lexical, syntax analysis, and symbol table generation.
Achieved code optimization and robust error handling.

Page Rank Algorithm
                  Implemented Map Reduce for PageRank, automating score calculations efficiently.
                
AccomplishmentsImplemented Map Reduce for PageRank on 2002 Google contest data.
Automated iterative PageRank score calculation in a distributed environment.
Leveraged Map and Reduce for iterative dataset processing.
Simulated PageRank's iterative nature effectively with Map Reduce.