BOMMALA ROHITH | 21ME31056

[Link].(Hons.) in MECHANICAL ENGINEERING and [Link]. in MANUFACTURING SCIENCE AND

ENGINEERING
MICRO SPL. in ARTIFICIAL INTELLIGENCE AND APPLICATIONS
EDUCATION
Year Degree/Exam Institute CGPA/Marks
2026 [Link] Dual Degree 5Y IIT Kharagpur 7.96 / 10
2021 Intermediate Public Examinations Sri Chaitanya Jr Kalasala 98.7%
2019 S.S.C Public Examinations Sri Chaitanya School 10.0 / 10.0
INTERNSHIPS
Machine Learning and Applied AI Intern | Kurious Kosmonauts Technology Services Pvt. Ltd OCT 2024 - DEC 2024
Objective: Built AI-driven trading system combining news sentiment and quantitative signals to enhance market forecasting accuracy and returns
•Processed 1M+ financial news articles using FinBERT NLP, creating sentiment features that improved model prediction accuracy by 12%
•Engineered a deep quantitative AI trading system using sentiment signals, boosting alpha returns 83% while cutting max drawdown 32%
•Cut RMSE by 57% deploying production XGBoost, LSTM &Transformer ensembles via FASTAPI, serving over 500 trading predictions daily
Data Science and AI Intern | IMEDTECH LLP MAY 2024 - JUN 2024
Objective: Built AI-driven system predicting multi-horizon stock/crypto returns with confidence intervals to optimize risk-adjusted portfolio decisions
•Built an end-to-end AI trading system with a Transformer model, achieving a 0.45 increase in Sharpe Ratio via probabilistic forecasting
•Mitigated model instability and overfitting using ensemble methods and Monte Carlo dropout, reducing maximum drawdown by 35%
•Deployed the AI model as a high-availability REST API, serving 1,000+ daily predictions with 99.8% uptime in a production environment
Data Science and AI Intern | IMEDTECH LLP JUN 2024 - JUL 2024
Objective: Built ML-powered anomaly detection system to predict industrial pump failures early, preventing costly downtime and productional losses
•Developed a high-precision ML anomaly detection system achieved a 97.3% detection rate, preventing costly industrial pump failures
•Engineered over 1,100 complex temporal features from 220K records, providing critical data foundation for the model's 99.57% ROC-AUC
•Deployed a real-time inference API with FastAPI, enabling 24/7 pump monitoring with a highly consistent 12ms low-latency response
Machine Learning & Applied AI team | Kurious Kosmonauts Technology Services Pvt. Ltd OCT 2024 - DEC 2024
Objective: Developed Graph Neural Network system for intelligent fraud detection in bank transactions by analyzing money transfer patterns
•Designed GraphSAGE fraud detection system analyzing 108K transactions across 199K accounts, achieving 98.8% F1 and 99.9% ROC-AUC
•Engineered 20 graph-based features from transaction behaviors, boosting fraud detection accuracy by 15%, reducing training time by 40%
•Effectively mitigated extreme 0.13% class imbalance using weighted loss to identify 99.5% of fraud cases with minimal false positives
PROJECTS
Explainable Deep Learning Model for Pneumonia Detection in Medical Imaging | Prof. Suman Chakraborty AUG 2024 - NOV 2024
•Built and deployed a ResNet50 model for pneumonia detection on chest X-rays achieving 98.7% accuracy, matching doctor-level trust
•Integrated Explainable AI to highlight abnormal regions on patient chest X-rays, improving clinical trust and explainability for 150+ scans
•Developed a hospital-verified platform optimized for low latency and efficiency enabling faster diagnosis for underprivileged patients
Scalable Churn Prediction Using Random Forest on Imbalanced Data | Self Project DEC 2024
•Built a Random Forest model on 193K+ records, handling severe class imbalance to achieve 82% precision in identifying at-risk users
•Engineered a reproducible ML workflow with ROI-aligned metrics, creating a scalable system for data-driven campaign performance
•Architected a tiered retention strategy from churn probabilities, enabling channel-specific interventions to retain high-value customers
Symptom-to-Disease Prediction using Bipartite Graph Neural Networks | Prof. Suman Chakraborty JAN 2025 - APR 2025
•Improved disease prediction accuracy by 25% using bipartite graphs with GraphSAGE, batch normalization, and regularization techniques
•Designed custom aggregation and deep classification layers to strengthen multi-symptom learning and improve model generalization
•Reduced training time by 90% through stratified sampling, early stopping, and adaptive learning rate scheduling for 731 imbalanced classes
Research Robot Restoration: Real-Time Control & Trajectory Optimization | Prof. D. K. Pratihar
•Engineered real-time control system for an 18+ DOF hexapod robot, integrating hardware and software to process motion data at 10Hz
•Applied optimization & algorithm design using inverse kinematics and nonlinear methods for motion planning &redundancy resolution
•Developed multi-threaded pipelines for sensor feedback and motor actuation, turning robotics research into production-ready code
SKILLS AND EXPERTISE
Skills: Bayesian Methods | C | C++ | Critical Thinking | Excel Advanced | FastAPI | Feature Engineering | Git/GitHub | MATLAB Basics |
Google Cloud Platform | NumPy | Pandas | Python | R | SQL | Streamlit | Tableau | Team Leadership &Mentoring | Web Scrapping
MOOCS [Be10X]: Generative AI Mastery | Corporate Communication | Prompt Engineering Course | AI for Coding | Data Analytics Using AI

COURSEWORK INFORMATION
Probability and Statistics | Linear Algebra | Multivariate Calculus | Numerical Methods | Optimization Techniques | Stochastic Processes
Data Structures and Algorithms | Machine Learning | Deep Learning | Natural Language Processing | Computer Vision | Business Analytics
EXTRA CURRICULAR ACTIVITIES
•Hockey Team Member, IIT KGP General Championship, demonstrated strategic gameplay, teamwork and high-pressure performance
•National Sports Organization certified trainer; spearheaded Anti-Corruption campaign, demonstrating social impact and ethical leadership
COMPETITION/CONFERENCE
Data Science Competition Winner | FIFS Sports Data Gameathon 1.0 | Top 10 National Finalist
•Engineered a Dockerized ML pipeline with Beautiful Soup for automated data processing, placing Top 10 nationally among 52 elite teams
•Developed a predictive XGBoost model using SMOTE to handle class imbalance and GridSearchCV for optimal hyperparameter tuning
•Presented research on AI in sports at WAVES Summit, securing a 10K prize and engaging with the Dream11 CEO on core industry strategy
!Self declared by the student, CDC could not verify the relevant documents