Pranav Gunhal_

CS + Education · UCSB · NLP & Education Equity Research

The best researchers and teachers are the ones who teach people, not just curriculum.

My overarching goal is to create AI that benefits structurally underserved students, whether they are transfer students navigating complex articulation pathways, diaspora students learning their mothertongues, or high schoolers stressed by college admissions.

I study CS + Education at UCSB. My research focus is the intersection of machine learning and education equity. I want to investigate not just whether AI systems perform well, but whether they can be trusted when deployed in high-stakes learning environments.

My software engineering background gives me the foundations to design new, inclusive systems, and I want to apply these skills in novel research contexts.

Course Articulation ML

First-generation and low-income students are disproportionately affected by the complex course articulation between California Community Colleges and UC campuses, especially for Engineering curriculum. I am designing and evaluating NLP pipelines (with TF-IDF, Sentence-Transformer Embeddings, and skip-gram models) on more than 1,000 course descriptions to explore semantic similarity between courses across institutions. Our preliminary findings suggest that sentence-transformer embeddings have high potential, which we aim to extend to minimize bias in articulation policies.

Earthquake Modeling & Prediction

CNN-LSTM models mapping Mercalli to Community Intensity using large geospatial datasets. Optimized model inference on HPC/GPU clusters with parallelization for latency-sensitive alerts.

Election Sentiment Analysis

Transformer-based sentiment classifiers for the Indic language with novel data augmentation to overcome labeled data scarcity. Presented at CSSE 2023.

Semantic Coherence · Mental Health

CNN-LSTM model assessing semantic coherence of Reddit posts as a predictive signal for mental health diagnostics. Explores how linguistic structure surfaces psychological state.

Indic Language Parsing

Novel algorithm converting Brahmic scripts into phonetic equivalents for downstream NLP. Enables analysis of Hindu religious texts with standard ML tooling.

Herbalyst

AI-powered cross-platform app delivering personalized herbal tea recommendations via GPT-4. Firebase backend, Razorpay payments, and a medication tracker developed under Dr. Corky Wicks.

UCSB Arch Lab · BACI Dataset

Tool suite for scraping and visualizing BACI export data, mapping supply chain logistics for critical minerals used in data centers. Collaborating with PhD students.

Software Engineer @ Ourfreedom.ai

• ▸Building Reddit-style community feeds for a social networking platform connecting incarcerated individuals' families — 2,000+ active users.
• ▸Implemented Sign in with Apple & Facebook OAuth, boosting onboarding rate by 35% and reducing login friction on iOS.
• ▸Reviewed 50+ PRs; led iOS production deployments and TestFlight builds with zero critical regressions across 3 versions.