Hi, I'm Sushanth!

I love exploring data, connecting the dots to uncover patterns, and building models that help achieve meaningful impact. At the heart of my work is a belief in the power of storytelling through data—seeing raw numbers weave a compelling story of their own.

Take a look around, and feel free to reach out!

🔹 Education 𓆝 𓆟 𓆞

✰ Manipal Institute of Technology

Bachelor of Technology, Computer Science Engineering (2017 - 2021)

Minor Specialization in Intelligent Systems

Coursework: Machine Learning, Artificial Intelligence, Distributed Cloud Computing, Computer Vision, Natural Language Processing, Social Network Analysis, Operating Systems, Data Structures and Algorithms

✰ Carnegie Mellon University

Master of Information Systems Management (2023 - 2024)

With a Focus in Business Intelligence and Data Analytics

Coursework: Machine Learning, Operationalizing Artificial Intelligence, Distributed Systems, Unstructured Data Analytics, Computer Vision, Advanced Business Analytics, Data Analytics with Tableau, Organizational Design and Implementation

Teaching Assistant: Applied Econometrics (Spring’24 & Fall ‘24) - Statistical Inference, Hypothesis Testing

🔹 Professional Experience 𓆝 𓆟 𓆞

✰ Machine Learning Engineer | Quantiphi Inc.

[Machine Learning (XGBoost, HMM, GMM, SciKitLearn), Data Engineering (PySpark, Azure Databricks, MLflow), Forecasting & Optimization, Cloud & MLOps (CI/CD, Model Deployment, Kubernetes), Data Analysis (SQL, Pandas)]

• Developed an Inventory Tracking System by leveraging advanced data analysis and machine learning models to detect product mismatches and optimize the Space-to-Sales ratio for retail stores. Recognized and awarded by the team for driving operational efficiency and improving data accuracy by 35%. - Presented the Inventory Tracking System at the annual portfolio meeting, translating analytical findings into actionable business solutions and driving consensus on inventory optimization strategies
• Developed and deployed an Out-of-Stock Detection System leveraging XGBoost for time-series forecasting and anomaly detection, enabling proactive inventory management for 30,000+ retail stores. Built scalable data pipelines using Python and PySpark, improving data ingestion, validation, and model performance
• Researched and experimented with Hidden Markov Models (HMM) and Gaussian Mixture Models (GMM) to enhance stock replenishment predictions, optimizing forecast accuracy by 1-3% across key performance indicators (KPIs).
• Led the migration of legacy machine learning pipelines to a scalable, cloud-based architecture using Azure Databricks, PySpark, and MLflow, ensuring seamless integration and improved model performance.
• Conducted research on pricing anomaly detection, customer segmentation, and labor demand forecasting, leveraging order and invoice datasets to identify pricing inconsistencies and optimize workforce allocation strategies for better operational planning

✰ Machine Learning Intern | Quantiphi Inc.

[Programming (Python), Machine Learning (SciKitLearn), Data Analysis (Pandas), Cloud Technologies (Microsoft Azure)]

• Optimized a Recommender System for Coca-Cola Bottlers by refining data pipelines, implementing Latent Factor Models and Random Forest for collaborative filtering and ranking, and optimizing model performance. Deployed on Azure Databricks, this system boosted product assortment accuracy, leading to a 21% increase in quarterly profit margins and streamlined inventory management across 30,000+ stores
• Developed an end-to-end Customer Churn prediction model with 96% accuracy, leveraging complex datasets and advanced machine learning algorithms

✰ Research Analyst Intern | Carnegie Mellon University

[Machine Learning (XGBoost, SciKitLearn, Time-Series Forecasting), Blockchain & Smart Contracts (Ethereum, Bitcoin), Data Engineering (PySpark, SQL), Predictive Modeling, Data Analysis (Pandas, NumPy)]

• Designed and developed a prototype utility token for the travel industry using the Ethereum blockchain platform, enabling secure and decentralized travel transactions
• Conducted market analysis on 100K+ transaction records to identify token adoption trends and forecasted a 20% growth potential in the tourism sector
• Researched blockchain security protocols, consensus mechanisms (PoS & PoW), and smart contract vulnerabilities to enhance the reliability and efficiency of decentralized transactions. Presented findings and technical insights to faculty and stakeholders

🔹 Leadership and Awards 𓆝 𓆟 𓆞

🏆 Awards

Outstanding Innovation Award – Recognized at Quantiphi for leading the development of the Inventory Tracking System, improving operational efficiency and data accuracy by 35%.

Contribution and Reliability Award – Awarded at Quantiphi for the successful deployment of an Out-of-Stock Detection System.

🌟 Leadership & Extracurriculars

Member, Data Science Club, Carnegie Mellon University – Engaged in discussions, mentorship, and workshops focused on AI and machine learning advancements.

Social Media Head, Revels (Manipal’s Annual Cultural Fest) – Led content strategy and execution, driving engagement for one of Manipal’s biggest cultural events.

Sub-Head of Writing, Manipal The Talk Network (MTTN) – Oversaw editorial direction, content planning, and writing mentorship.

Editor, Editorial Board – Managed and refined high-quality written content for university publications.

Writer, MTTN – Contributed articles covering technology, culture, and student affairs, amplifying student voices through impactful storytelling.

🔹 Projects 𓆝 𓆟 𓆞

TCS Road Intelligence Project

Leveraging AI for Smarter Roads

Code Repository

Developed a road intelligence system to enhance traffic management and road safety using AI-based image recognition and data analytics. The project involved detecting and analyzing traffic patterns, road conditions, and potential hazards to optimize real-time decision-making for transportation agencies.

Key Technologies:

• TensorFlow
• OpenCV
• Python
• Keras
• Data Analytics

LLM Efficiency Enhancement: Speculative Decoding Model Generation

Accelerating Large Language Model Inference with Speculative Decoding

Code Repository

This project focuses on improving the inference speed of Large Language Models (LLMs) by leveraging speculative decoding. We created a smaller draft model using traditional model compression techniques, such as pruning, quantization, and layer compression, to generate tokens faster. The target LLM then verifies these tokens in parallel, resulting in a significant speedup in the decoding process.

Key Technologies:

• PyTorch
• Large Language Models
• Model Compression Techniques
• Speculative Decoding
• Deep Learning Optimization
• AI Model Efficiency

Pennsylvania Elections Analysis

Analyzing Election Trends with Tableau and SQL

Code Repository

This project focuses on analyzing Pennsylvania’s election data to uncover trends and insights related to voter demographics, election outcomes, and precinct performance. Using Tableau Desktop for data visualization and SQL for querying the dataset, we created interactive dashboards that provide a comprehensive view of voting patterns across different regions.

Key Technologies:

• Tableau Desktop
• SQL
• Data Visualization
• Data Analysis

Traffic Prediction Model

Predicting Traffic Patterns using METR-LA Dataset

Code Repository

This project involves developing, deploying, and monitoring a traffic prediction model using the METR-LA dataset. We train LSTM and GRU models to predict traffic patterns and deploy the model using Docker and Kubernetes. The model’s performance is then monitored using Evidently, ensuring its reliability in real-world scenarios.

Key Technologies:

• LSTM
• GRU
• Docker
• Kubernetes
• Kubeflow Pipelines
• Evidently
• Flask

Loan Lending Case Study

Analyzing Loan Default Risks using Machine Learning

Code Repository

This project explores loan data to identify key factors influencing loan defaults. We leverage machine learning models to predict loan repayment outcomes, enabling financial institutions to make data-driven lending decisions. The project includes data preprocessing, feature engineering, model training, evaluation, and deployment.

Key Technologies

• Machine Learning Models: Logistic Regression, Random Forest, XGBoost
• Deployment & Infrastructure: Flask, Docker, Kubernetes
• Data Processing & Storage: SQL, Scikit-learn
• Visualization: Matplotlib, Seaborn

USD Exchange Rate Optimizer

Optimizing USD Exchange Rate Conversions using Machine Learning

Code Repository

This project focuses on predicting and optimizing USD exchange rates using historical data and machine learning techniques. By analyzing past trends and market indicators, we aim to provide accurate exchange rate forecasts, helping users make informed currency exchange decisions. The system is deployed as a web service, enabling real-time exchange rate predictions.

Key Technologies

• Machine Learning Models: Time Series Forecasting, LSTM, ARIMA
• Deployment & Infrastructure: Flask, Docker, Kubernetes
• Data Processing & APIs: Pandas, NumPy, Forex APIs
• Visualization: Matplotlib, Seaborn

🔹 Volunteering 𓆝 𓆟 𓆞

• Content Creator at Volunteer for India – Created engaging and informative content to support social initiatives.

• PLUS Tutor at Carnegie Mellon University – Tutored middle school students in math, helping them build a strong foundation in the subject.