The Beginning
Present Day
Senior · Research & CS · Granite Bay High School
"A picture is worth 1000 words"
At a Glance
For years, I treated effort like a contract: set a goal, work hard enough, and the result would follow.
After six months and 150 hours of squash practice, I expected to earn a Gold Division shirt. Then I lost the match that would have earned it.
The loss made pressure physical. My palms began to sweat before courts, podiums, and classrooms where the outcome might reveal something about me.
After the loss, I found ways to lower the stakes: the shared pressure of volleyball and the back of the room during hackathon presentations.
At UC Irvine’s GSET program, there was nowhere to hide. I memorized my part, walked onstage, and froze halfway through the presentation.
I had prepared obsessively. Still, the outcome I expected never arrived.
My mom brought me back to a line from the Bhagavad Gita: I had a right to my actions, not to their fruits. At first, that sounded like an argument against ambition.
Eventually, I understood the distinction. Aspiration asks me to prepare fully. Attachment says the work is worthless unless it produces the exact result I imagined.
My preparation, choices, and response afterward were mine. The score, applause, and final outcome never fully were.
I tested that idea at a Tech4Silvers session for 30 seniors at a local center. My palms dampened around the microphone, but I delivered the workshop and stayed for every question.
The goal was not to prove that I felt no fear. It was to help the people in front of me. By the next session, 50 more seniors wanted to join.
Standing up became a choice to remain useful while feeling uncertain.
During a timeout against Jesuit High School, I noticed my teammate Colin’s hand shaking as he waited to be subbed in. I recognized the feeling immediately.
I told him, “You got this, bro,” and meant it. We won that game and later won the state championship.
My palms still dampen before games and presentations. They just do not decide whether I step forward anymore.
I still set ambitious goals and prepare hard. The difference is that I no longer treat an outcome as proof that the work mattered. I can own my choices without pretending to own every result.
That mindset now shapes how I approach research setbacks and service and competition.
Underneath the work
The threads that run underneath everything else on this page.
Technology & Innovation
What I'm Working Toward
Service & CommunityService has required me to speak up even when I am nervous. I have led technology workshops for seniors and presented the results of our clothing drive at Rancho Cordova City Hall. Those experiences taught me that community work depends on showing up, organizing the details, and explaining why the work matters.
My recent work has pulled me toward AI and machine learning, especially questions about how we evaluate intelligent systems and who those systems work well for. I have explored behavioral consistency in AI agents, fairer AI-text detection, and computer-vision support for a serious game. What interests me most is moving between research and building: defining a useful question, testing it with real data, and turning the result into something people can use.
Right now, I am focused on becoming a stronger researcher and builder. I want to finish and publish the projects already underway, keep developing more rigorous ways to evaluate AI systems, and continue building technology for health, education, and civic access. I also want to stay open to new collaborators and problems instead of forcing everything into one narrow path.
Academic foundation
Graduating from Granite Bay High School in 2027.
Class standing
1st
Ranked 1st of 504
Weighted GPA
4.69
4.0 Unweighted
SAT Superscore
1530
790 Math • 740 Reading & Writing
Beyond the classroom
Outside class, I lead community service projects, compete in programming and volleyball, and perform violin. These are the teams, clubs, and organizations where I spend most of my time.
A free fellowship co-founded to give underprivileged high schoolers (including QuestBridge scholars) who lack access to scientific research mentorship a path into AI research. Over a four-month program, delivered weekly skill-building lectures and paper-review sessions for 20+ students across 4 teams, guiding them from an initial idea to a submitted paper.
When I joined “Coding Club” as a freshman, it met once a week to casually solve a few problems before HPE CodeWars. As President, I’ve rebuilt it into the Competitive Programming Olympiad Club. I design weekly practices on algorithms and data structures, teach Python/Java solutions live, and organize our teams for all four USACO contests, NASA Space Apps, HPE CodeWars, and other hackathons. Under this new structure, our members earned a NASA Space Apps Global Nominee, multiple CodeWars trophies, and a big jump in USACO participation. At a school where most STEM kids lean pre-med, this club has become the main hub for students who want to push themselves in CS and competitive coding.
NHS is the main service hub at my school, and as Sergeant at Arms I help plan and run projects instead of just showing up to volunteer hours. I’ve helped coordinate drives under our PraNam initiative, including a warm-clothes collection that delivered 500+ items to students at Kinney High School and a shoe drive that shipped over 150 pairs to families in Africa. I also help manage meeting logistics and keep events running smoothly. NHS has been where I practice turning big, vague goals like "help the community" into specific, organized events that actually deliver.
I’ve been programming for six years and competing for three. I qualified for the USACO Gold division and train year-round by solving past contests in Python and Java, focusing on graph algorithms, DP, and implementation speed. Outside USACO, I’ve led or joined teams for events like NASA Space Apps, HPE CodeWars, and the YoungWonks National Coding Challenge, where we’ve earned a Global Nominee in Space Apps and a 2nd-place regional finish at CodeWars. These contests force me to write clean, efficient code under time pressure and to communicate clearly with teammates when the clock is ticking.
At UCI x GATI’s GSET program, I spent a summer studying how game design can support education and behavior. Working in Unity with C#, my team built a prototype that combined reward schedules, difficulty curves, and machine-learning-driven customization. We wrote a lightweight business plan and presented the demo to a panel of faculty and judges. The project later informed my work on Flora Frontier.
At UCLA COSMOS, I worked with Prof. HTB to study whether AI theorem provers use stable reasoning strategies across different problems, rather than only whether they reach the right answer. I represented Lean proofs as fingerprints of their tactics, structure, and errors, then used machine learning and similarity analysis to compare human and AI proof behavior. The work analyzed 260,103 proofs, validated the fingerprints at 0.923 macro F1, and tested the findings across six robustness checks.
At a glance
I started Tech4Silvers after realizing many seniors in our community were missing digital wildfire and emergency alerts. It started with one workshop and has grown into a student-run organization and GBHS club that partners with senior centers across Sacramento and Northern California. I recruit and train volunteers, design slides and handouts, and lead sessions on smartphones, scams and phishing, online safety, telehealth, and local alert systems. We’ve helped dozens of seniors set up alerts, avoid fraud, and reconnect with family over video calls. The work has taught me how to organize community programs around practical needs.
My brother Naman and I started PraNam in the summer of 2022 and built it into a network of collection drives, community partners, and practical service projects. We have collected and donated more than 1,000 items, including 500+ warm clothing items, 600+ eyeframes, and 100+ educational books. Working with the Rotary E-Club of Silicon Valley, I also led a 150-pair shoe drive for families in Africa and a warm-clothes drive for students at Kinney High School in Rancho Cordova. That local drive received mayoral recognition, and I spoke about the project at Rancho Cordova City Hall.
At a glance
I have played club volleyball for nine years and school volleyball throughout high school. As an outside hitter, I helped club teams earn silver at USAV Nationals (14 Open), gold at the SCVA SoCal Cup, gold in the NCVA Power League, and multiple qualification medals. At Granite Bay, I played JV as a freshman, winning league and the Stockton JV tournament, then joined varsity as a sophomore. Moving between club and school teams taught me to reset after mistakes, contribute in different roles, and balance long practices, film, and travel with a demanding academic schedule.
I’ve played violin for five years, starting in school orchestra and later joining the Sacramento Youth Symphony. Between rehearsals, sectionals, and concerts, I’ve performed in over fifty events across California. Playing in an ensemble taught me to listen carefully, adjust my intonation, match phrasing, and stay with the conductor. Violin gives me a steady routine outside my technical work.
Research, systems, and impact
This is the 3 AM version of me from the story above, six years later. Almost none of it started as a research plan. It started as a problem that bothered me and wouldn't leave. So instead of splitting these into “projects” and “research,” they're grouped by what they're actually about.
Correcting Euler buckling predictions without leaking test data
Research Author · Johns Hopkins University · 6 hours/week
The first version of this project used cross-validation that leaked information between train and test sets, producing results that looked far stronger than they really were, and was rejected by every venue it was submitted to for superficial novelty.
Ran 147 physical buckling tests on dried pasta strands (7 lengths × 4 diameters) to test where Euler's 250-year-old buckling formula breaks down for non-ideal materials. Rebuilt the entire analysis from scratch with validation that keeps the test data genuinely unseen. The successor manuscript was accepted for publication in IEEE Access in August 2026. Its central result is that direct machine learning performs worse than the classical formula, while a correction layer built on top of that formula cuts the remaining error by about 40% and holds up on strand thicknesses it was never trained on.
Accepted for publication in IEEE Access, August 2026
Mentorship
Dr. AR
UConn; Associate Professor at Polytechnic University, Orlando
Paper
A Leakage-Safe Hybrid Physics-Machine Learning Framework for Correcting Euler Buckling Predictions in Non-Ideal Columns
Founder & President
I started Tech4Silvers after realizing many seniors in our community were missing digital wildfire and emergency alerts.
Explore the Tech4Silvers storyMaking AI reasoning traceable, and measuring whether agents stay consistent
Research Assistant · Massachusetts Institute of Technology (remote)
Cold-emailed Prof. MK at MIT about research on making AI reasoning traceable to real data.
Joined as a remote research assistant, contributing production features across dozens of merged pull requests to a visualization platform and API, expanding real-time 3D map rendering, collaboration tools, and backend reliability.
Mentorship
Prof. MK
MIT
Paper
Research on measuring behavioral consistency in AI agents
Cutting the ESL false-accusation rate from 29% to 3.9%
Lead Researcher
Built a dual-stream AI-text detector after learning that commercial AI detectors systematically misclassify writing by English-as-a-Second-Language (ESL) speakers as machine-generated.
Combines a Native-Expert stream (sub-word perplexity) with an ESL-Expert stream (character-level LSTM trained on learner corpora), grounded in second-language-acquisition theory.
