Raghav Dheri Raghav Dheri

On the first project, I developed multiple web scrapers in Python using Selenium and BeautifulSoup to automate data collection and PDF content extraction from various multilingual websites. I also used OpenPyXL to organize the scraped data into Excel reports.

For the second project, I led the development of a Retrieval-Augmented Generation (RAG) based Financial Recommendation System using Python and Ollama LLM. Users can input financial data, get concise summaries of key insights, ask context-aware questions, and receive results powered by vector-indexed retrieval, along with auto-generated tables and visualizations for clearer insights.

I integrated IMU sensor data into ML algorithms and built an iOS gesture-recognition application for visually impaired users, achieving 92% accuracy across all supported gestures. The goal was to make the phone navigable without looking at the screen.

I then collaborated with the therapy department to run a 12-participant user study, which helped us refine the model further. Retraining on the expanded dataset brought false positives down from 16 per hour to under 4.

I led weekly recitation sessions for 25+ students, running targeted problem-solving workshops and holding 1:1 office hours for anyone stuck on vector fields, Lagrange multipliers, or the usual suspects. I also worked with faculty on designing assessments and grading exams, keeping evaluation consistent across course sections.

One of the things I enjoyed most was building visual learning aids for the 3D concepts that trip most students up. Multivariable optimization and vector analysis get a lot easier once you can see them. My section's average grade ended up about 15% higher than prior semesters.

My main project was a distributed payment processing system in Java and Spring Boot. I worked on the microservice decomposition that let us handle 3,000+ requests per second, and used async processing and backpressure to cut transaction latency by around 15% across the three services we touched.

On the infrastructure side, I set up a Jenkins CI/CD pipeline from scratch with automated unit testing, which raised test coverage from 60% to 88% and meaningfully shortened our merge-to-production cycle.

Co-led a 5-person team building a Chrome extension that integrates Gemini AI into students' workflows. The idea was to surface instant AI-powered insights wherever they were already working, without the context switch of opening another tab.

I focused on the full-stack architecture using React, Tailwind CSS, and the Chrome APIs, and shipped the context-menu and side-panel functionality end-to-end. Integrating the Gemini API came with its own performance and responsiveness challenges, which made me a lot more careful about writing clean code.

I led weekly algorithm workshops for 35+ students, covering dynamic programming, graph algorithms, tree traversals, and complexity analysis. Basically everything they'd actually see in a tech interview. I redesigned the recitation around live coding and interactive labs rather than lecture-style review, designing coding exercises around binary search trees, shortest paths, and recursion.

By the end of the semester, attendance was up 20% and average quiz scores were up 30%.

I tutored 7 students weekly in one-on-one sessions across linear algebra and calculus II, adapting how I explained things (differential equations, linear transformations, eigenvalues) to each person's learning style and background. Some students wanted intuition first. Others wanted to drill problems.

Across the cohort, students saw about a 40% average improvement in their course performance. Watching someone go from stuck to actually getting it is one of the most rewarding things I've done in college.

I led a 9-person council serving 300+ residents. My responsibilities included planning community events, allocating a $1,240 annual budget with full P&L responsibility, and running the executive meetings. We focused on data-driven programming, which bumped engagement up about 30% over the previous year.

It was my first time managing a team of that size, and most of what I now know about running a small group (setting agendas, chairing meetings, making decisions quickly) came from this role.