Intro

Welcome! I'm Shashank Muthuraj, an AI/ML Engineer and a Master of Science in Software Engineering candidate at San Jose State University. Originating from a strong foundation in Computer Science with a Bachelor of Technology from PES University, Bangalore, I am passionate about deep learning, machine learning, and transforming complex ideas into innovative solutions. My technical proficiency spans Python, PyTorch, and various other ML technologies, focused particularly in deep learning and natural language processing. Dive into my portfolio to see the projects I've worked on and the technologies I've mastered.

For a closer look at my work and the impact of my projects, please check out my projects section.

Work Experience

May 2024 - Dec 2024

Machine Learning Engineer Intern, KanTime, Plano, TX

I led the development of a distributed data pipeline using PySpark to process over 500,000 patient records, resulting in a 30% increase in ingestion speed and a 25% reduction in data processing costs. I also owned the end-to-end fine-tuning of ClinicalBERT with PyTorch and QLoRA techniques for multilabel classification of ICD-10 codes, which improved model accuracy by 40% and reduced manual medical coding efforts by approximately 15 hours per week. Additionally, I developed LLM-based automation solutions for medical coding, significantly enhancing diagnosis prediction and streamlining patient onboarding workflows.

Jul 2022 - Jul 2023

Software Engineer, SAP Labs, Bangalore, India

I spearheaded backend development for SAP Joule, driving a 20% improvement in chatbot operational efficiency by implementing machine learning-based intent and entity classification, integrated with technologies such as Java, Spring Boot, Hana DB, and Kafka. I optimized the microservices architecture to handle over 1 million monthly requests, working closely with cross-functional teams in Scrum sprints to refine conversational flows and enhance user satisfaction by 30%. Additionally, I built and automated CI/CD pipelines using Docker and Kubernetes, which shortened release cycles by 40% while ensuring 99.9% system uptime in production environments.

Jan 2022 - Jun 2022

Data Scientist, SAP Labs Bangalore, India

I implemented the DBSCAN algorithm to enhance point-based anomaly detection, which led to a 15% improvement in system monitoring accuracy within the SAP Focused RUN environment, thereby strengthening overall reliability. Working alongside ML engineers, I helped transition prototypes into deployable models by standardizing preprocessing pipelines, containerizing training workflows, and ensuring reproducibility across staging environments. Additionally, I developed an evaluation framework that incorporated rule-based labeling and domain expert feedback to validate clustering results, enabling iterative tuning of DBSCAN parameters and improving anomaly detection efficiency.

For a more in-depth look at my projects and technical contributions, check out the projects section.

Projects

Here are some of the key projects that showcase my expertise in AI and ML technologies:

SwiftSelect: AI-powered Recruitment Platform

Designed an AI-driven recruitment platform leveraging Semantic Textual Similarity (STS) and transformer-based embeddings for accurate candidate-job matching. Open-sourced a fine-tuned LoRA adapter model on HuggingFace with over 1,000 downloads, enabling efficient semantic matching between resumes and job descriptions. Deployed real-time inference models on AWS SageMaker using FastAPI and EKS, implementing model quantization to achieve sub-100ms response times at scale. Orchestrated an end-to-end MLOps pipeline with automated training, evaluation, and deployment using SageMaker Pipelines, incorporating CI/CD and A/B testing frameworks.

Visa Support Virtual Assistant (RAG)

Developed a virtual assistant for F1 and H1B visa applicants using a Retrieval-Augmented Generation (RAG) framework to provide context-aware responses, reducing manual query resolution time by 40%. Leveraged FAISS indexing with LLaMA 3.1-8B and Mistral 7B for document retrieval, achieving a 98% recall rate on a 30,000-word corpus of official documents. Fine-tuned both models with Direct Preference Optimization (DPO) to improve policy adherence and reduce visa-related inaccuracies. Optimized inference pipelines for speed and reliability in production.

AI-Driven Foot Traffic Monitoring: A New Era of Retail Insight

Developed an advanced AI system to analyze retail foot traffic using TensorFlow, OpenCV, YOLOv5, and Python. This solution was integrated with existing CCTV systems to enable real-time monitoring and data analysis, complete with a web interface built using HTML, CSS, and JavaScript for insightful analytics. The project emphasized accuracy, scalability, and cost-effectiveness, revolutionizing traditional retail foot traffic analysis methods.

Cyberbullying Detection System

Pioneered a sophisticated cyberbullying detection system utilizing advanced machine learning models like RoBERTa, SVM, Ridge Classification, and XGBoost. The system achieves high accuracy in identifying instances of cyberbullying across various categories, demonstrating its robustness and potential for social media integration to augment online safety.

Predictive Analytics of Economic Factors

Analyzed the effect of economic patterns and their path towards stability, especially during the pandemic, to enhance the decision-making abilities of business groups. Employed a combined model including Linear Regression, the Winter-Holtz method, SARIMA, and LSTM to provide valuable and accurate insights.

Contact

I'm always excited to connect with fellow enthusiasts, professionals, and anyone passionate about Machine Learning and technology. Whether you have a question, want to collaborate on a project, or just want to chat, feel free to reach out! You can contact me through the form below or connect with me on LinkedIn. Looking forward to hearing from you!