Priktish Rao Suntoo

I build machine learning and GenAI systems with a research mindset—turning messy data into validated models and reliable user-facing applications. I focus on reproducibility, clear evaluation, and deployment-ready pipelines.

In my applied work, I engineered a League of Legends match predictor using PyTorch and scikit-learn, training a logistic regression model on 9,879 ranked matches and reaching ~74–75% test accuracy. I’ve also built RAG experiences that retrieve source-cited answers and support real multi-turn workflows.

My PDF QA Bot ingests arbitrary PDFs, generates embeddings, and serves citations through the Claude API, with a Gradio interface for end-to-end interaction. I extended an IBM final project by integrating the Anthropic API, updating LangChain conventions, and implementing persistent multi-turn chat history.

Before completing the IBM course, I independently built Writhat, a full-stack AI academic writing platform. It features a Firebase Cloud Functions backend, AI-powered writing analysis via the Claude API, LaTeX equation rendering with MathJax, PDF summarisation, and multi-format citation management (APA/MLA/Chicago/IEEE) through the CrossRef API. Writhat demonstrates my ability to design and ship a complete application end to end, from backend architecture to frontend user experience, without external guidance.

As a Graduate Teaching Assistant and Research Assistant, I’ve taught 100+ undergraduates and developed mathematical models for superconductivity in topological flat bands—iteratively validating outputs against theoretical constraints. Earlier, I supported campus infrastructure as an IT Student Assistant across Windows, macOS, and Linux.