Saket Gudimella

Built a merchant-centric product recommendation engine for Clover POS using GNNs, improving precision by 14% via custom cohort-level train/test splits and optimizing Hit Rate & Recall@5. Developed NER pipelines including an MOA Entity Extractor (spaCy + rules) and Merchant Name normalization (regex, fuzzy matching, external data), reducing false positives by 20% through iterative tuning.

Developed deep learning super-resolution models (MIRNet, SwinIR) to enhance OHRC lunar images by 4×–8×, improving crater and boulder visibility. Fine-tuned with Charbonnier loss, benchmarked using PSNR & SSIM, and compared CNN/Transformer approaches (HBPN, DBPN, SwinFIR). Proposed a MIRNet-GAN hybrid to reduce blurriness and enhance fine lunar surface structures.

Built a disease prediction model using Llama 2 (7B) with LoRA and 4-bit quantization for efficient deployment on Intel AI platforms. Developed a symptom-based chatbot to improve healthcare accessibility, validated performance with real-time data handling and ML analysis, and enhanced prediction accuracy for impactful AI-driven healthcare solutions.

Developed real-time data workflows improving efficiency by 10%, analyzed machine data for predictive maintenance, and gained expertise in visualization tools (Google Data Studio, Power BI, Tableau). Presented insights that shaped project strategies and proposed an image-processing project for concrete quality testing to enhance assurance processes.

GPA: 9.21/10.00