🟢 Open to work
I am a freshman studying Computer Science, building foundational skills and early projects as I prepare to enter the field.
A selection of class assignments and personal projects that show my growing skills in programming, problem solving, and software fundamentals.
Personal Project · Sep 2026
DOCKET is a civic information platform designed to help Fremont residents find reliable local information in one place. It combines civic records, local news, live incidents, places, community posts, and issue voting into a single user experience. The platform uses Next.js, React, TypeScript, FastAP
Personal Project · Sep 2026
Reduced npm supply-chain analysis to 3 core operations - graph traversal, semver resolution, and OSV correlation - by modeling transitive dependencies as an actionable graph. Extended the system into deployed monitoring and API surfaces with signed webhooks.
Prevented unverified production changes by executing proposals against a shadow database, verifying checksums, supporting rollback, and withholding approval until proof passes. Reached 347 documented passing tests and zero production-write tools for agents by adding policy gates and tamper-evident l
Expanded cache-hit coverage beyond exact text matches by combining embeddings, cosine similarity, and SHA-256 identity.
I am a freshman studying Computer Science, building foundational skills and early projects as I prepare to enter the field.
I am Rohit Maruri, a first-year Computer Science student pursuing a bachelor's degree. I am building projects and skills in software and computing.
Experience
Undergraduate Researcher
Undergraduate Researcher | AWS Trainium and Neuron Research · Ongoing
Founder / Product Lead
Vivedly Labs Inc.
Software Engineer Intern
The Clarion
IT Assistant
Dominion Academy of Dayton
Education
San Francisco Bay University
B.S. Computer Science
Skills
Personal Project · Sep 2026
DOCKET is a civic information platform designed to help Fremont residents find reliable local information in one place. It combines civic records, local news, live incidents, places, community posts, and issue voting into a single user experience. The platform uses Next.js, React, TypeScript, FastAP
DOCKET is a civic information platform designed to help Fremont residents find reliable local information in one place. It combines civic records, local news, live incidents, places, community posts, and issue voting into a single user experience. The platform uses Next.js, React, TypeScript, FastAPI, Python, Amazon Bedrock AgentCore, Aurora DSQL, S3 Vectors, and BM25 retrieval.
Designed and built the full-stack application from the ground up. Created user workflows for browsing civic information and community issues. Built backend APIs with FastAPI and Python. Designed the data model and connected multiple information sources. Implemented semantic and keyword retrieval using S3 Vectors and BM25. Added source-grounded responses and citation verification.
DOCKET turned fragmented civic information into one organized platform. The project helped me learn how to build a product around real user needs while balancing frontend usability, backend design, data quality, retrieval accuracy, and AI reliability. The most important outcome was creating a system where users can not only receive information, but also understand where that information came from.
Personal Project · Sep 2026
Reduced npm supply-chain analysis to 3 core operations - graph traversal, semver resolution, and OSV correlation - by modeling transitive dependencies as an actionable graph. Extended the system into deployed monitoring and API surfaces with signed webhooks.
Reduced npm supply-chain analysis to 3 core operations - graph traversal, semver resolution, and OSV correlation - by modeling transitive dependencies as an actionable graph. Extended the system into deployed monitoring and API surfaces with signed webhooks.
Prevented unverified production changes by executing proposals against a shadow database, verifying checksums, supporting rollback, and withholding approval until proof passes. Reached 347 documented passing tests and zero production-write tools for agents by adding policy gates and tamper-evident l
Expanded cache-hit coverage beyond exact text matches by combining embeddings, cosine similarity, and SHA-256 identity.