I'm a graduate student pursuing an MS in Artificial Intelligence at Rochester Institute of Technology (Class of 2026), with a strong foundation in AI and Data Science from my BTech at SR Gudlavalleru Engineering College.

• Machine Learning, Deep Learning & Generative AI
• Brain-Computer Interfaces (BCI), EEG Decoding, and Cognitive Computing
• Few-Shot and Cross-Domain Learning for Satellite Imagery
• AI for Healthcare (Medical Imaging, EHR)
• AI Robustness & Interpretability (Adversarial Attacks, OOD Detection, SHAP, LIME)
• Neuro-symbolic Reasoning & Multimodal Transformers
• MLOps, Model Optimization & Scalable AI Deployment (Vertex AI, AWS)

• Advanced AI robustness techniques (e.g., adversarial attacks, OOD detection)
• Scalable model deployment using Vertex AI
• Multimodal learning and neuro-symbolic AI for real-world reasoning

• Research or open-source projects in ML/CV/NLP
• Projects applying AI in healthcare or environmental monitoring
• Exploratory work in neuro-symbolic AI or multimodal transformers

Pronouns: She/Her
⚡ Fun fact: I’m obsessed with stress-testing models—whether it’s causal patch injection or conceptual fragility metrics, I like to break models to make them better.

Languages: Python, R, SQL, Java, C, C++, JavaScript
ML/AI: Transformers (BERT, ViT, LLaMA, SAM), GANs, SHAP, LIME, RL, XGBoost, Bayesian Optimization, AutoML
Computer Vision & NLP: YOLO, Detectron2, VQA, RAG, NER, Image Segmentation, OCR, Grad-CAM, HuggingFace
MLOps & Cloud: Vertex AI, AWS, Docker, Kubernetes, MLflow, CI/CD, Streamlit, A/B Testing
Analytics & Tools: Tableau, Power BI, MySQL, MongoDB, Apache Spark, DVC
Frameworks & Libraries: PyTorch, TensorFlow, Keras, OpenCV, Scikit-learn, Flask, FastAPI, LangChain

• Developed a temporal convolutional decoder for real-time BCI EEG decoding using motor imagery datasets.
• Integrated FAISS retrieval to find semantically similar brain states, boosting decoding robustness.
• Built an interactive Streamlit UI for EEG tracking and simulated BCI control loop.

• Implemented contrastive pretraining + domain adaptation for urban/rural satellite imagery.
• Used ViT distillation + structured pruning to build efficient, lightweight classifiers for edge devices.
• Outperformed ProtoNet & MAML baselines by reducing inter-domain error by 12%.

• Designed a multimodal transformer to fuse EHR + time-series vitals.
• Applied SHAP/LIME for clinical interpretability and visualized temporal attention for diagnosis.
• Boosted precision to 91% (LogReg: 70%) for early cardiac risk assessment.

• Proposed Conceptual Fragility Index (CFI) to quantify CNN vulnerability to semantic OOD patches.
• Injected high-contrast CLIP-selected patches into Grad-CAM regions, degrading ResNet-18 accuracy by 40%.
• Automated the full CCS pipeline from patching to evaluation.

• Combined CLIP visual grounding with LLaMA-2 action planning for form-filling & task automation.
• Enabled multi-step reasoning via UI detection and natural instruction parsing.

• Built a semantic search + RAG system over 5K+ grant descriptions using OpenAI embeddings + FAISS.
• Achieved 28% improvement in Top-5 grant match accuracy using prompt tuning and reranking.

• Amazon: ML Summer School 2023
• NVIDIA: Generative AI for Diffusion Models
• NVIDIA: Fundamentals of CUDA Python

• Build 20+ AI research projects spanning BCI, vision, NLP, and robustness
• Contribute to 3+ open-source AI interpretability or MLOps toolkits
• Publish papers on adversarial robustness or cross-domain generalization
• Master scalable AI deployment with Vertex AI, LangChain, and neuro-symbolic reasoning stacks