Skip to content
Field Note 01

Building proof-based AI systems for students, founders, and technical teams.

I'm Advik Bhatt, a Rutgers CS + Data Science student building systems that turn scattered work into structured proof, strategy, and execution.

Founder, RolemateBuilder, FrenzyUSACS Tech DirectorRutgers CS + Data ScienceVoya IT InternMulti-Agent SwarmsRAG + Vector SearchLLM OrchestrationAI Systems Builder
Signal Strip
  • No. 01Rutgers CS + Data Science
  • No. 02GPA 4.0
  • No. 03Voya Python ETL automation
  • No. 04Voya Power BI executive visualization
  • No. 057 employees trained on Copilot-assisted automation
  • No. 06GMS Venture Lab: 8 teams guided to MVP
  • No. 07USACS: UHACCS 2027
  • No. 08DSC: 2 datathons planned
  • No. 09VibeCheck: AI-assisted coding guardrail
  • No. 10Frenzy: multi-agent evaluation platform
  • No. 1110 years saxophone and ensemble performance
Featured Systems

Case files

Three product systems, each with its own audience, inputs, outputs, and proof. Open a card to see the full file.

Case 01

Rolemate

Building toward Rutgers beta

AI proof-of-skill engine for student recruiting.

Rolemate helps students turn scattered proof of skill into job-specific application strategy. It maps projects, GitHub, club work, internships, resumes, and job postings into a structured proof map that shows what a student can credibly claim, what proof is missing, and what to build or say next.

Inputs

  • Job postings
  • Resumes
  • GitHub repositories
  • User documents
  • LinkedIn and networking context

Outputs

  • Capability maps
  • Requirement matches
  • Project recommendations
  • Referral strategy
  • Outreach drafts
  • Interview prep
  • Application strategy

Stack

Next.jsReactOpenRouterSupabasePostgreSQLRLSVector Search

Proof

  • Retrieval-augmented LLM orchestration with secure personalized retrieval
  • Multi-agent project recommendation swarm with Skill Gap, Resume Signal, Technical Depth, Portfolio Differentiation, Recruiter Interpretation, Execution Feasibility, and Synthesis agents
  • LinkedIn OAuth and referral-ranking surface in progress
  • Chrome extension job-posting capture in progress
  • Built around proof-of-skill instead of generic resume advice
Case 02

Frenzy

Production hardening

Multi-agent startup pitch pressure-testing platform.

Frenzy pressure-tests startup pitches, product concepts, strategy memos, and grant proposals through specialized Market, Product, Distribution, Technical, Skeptic, and Synthesis evaluator agents.

Inputs

  • Startup pitches
  • Product concepts
  • Strategy memos
  • Grant proposals
  • Founder Q&A answers

Outputs

  • Scorecards
  • Pitch rewrites
  • Investor objections
  • Validation roadmaps
  • Diligence memos
  • Structured JSON reports

Stack

JavaScriptNode.jsVercel Serverless FunctionsOpenAI APISupabaseStripeBYOK

Proof

  • Local readiness scoring and missing-element detection
  • Timed founder Q&A with evaluator deliberation
  • Structured JSON report generation with scorecards, rewrites, objections, validation roadmaps, and diligence memos
  • Server-side LLM orchestration and access-code validation
  • Production roadmap around Supabase Auth, PostgreSQL, RLS, Stripe Checkout, usage quotas, report credits, saved sessions, team spaces, and BYOK
Case 03

VibeCheck

Published and expanding

Python CLI guardrail for AI-assisted coding.

VibeCheck intercepts proposed AI-generated code changes before integration and checks whether the user understands why the change works.

Inputs

  • AI-generated code changes
  • Code context
  • User explanations

Outputs

  • Gate decisions
  • Adaptive technical questions
  • Competence evidence
  • Concept graph updates

Stack

PythonOpenRouterLangChainGitHub ActionsPyPI

Proof

  • Published Python CLI integrated with Claude Code development
  • Structured gate decisions with adaptive follow-up questions
  • Concept graph and local evidence model tracking competence over time
  • Planned PostgreSQL and pgvector layer for longitudinal competence history
Archive Drawer

  • CrewCast

    Multi-user staffing platform for temporary event labor, centralizing job distribution, crew coordination, and shift management.

    TypeScriptReact NativeNode.jsPostgreSQLSupabase

  • Neura

    AI-assisted intake platform for making early mental health information gathering more structured before provider review.

    JavaScriptReactNode.js

  • RUWorkingOut

    Gym-comparison web app for Rutgers students comparing gyms by location, timing, and traffic patterns.

    JavaScriptReactNode.jsGoogle Maps API
Operating Surface

Field log

Where I work, what I own, and what gets produced. Programs and products that students, founders, and recruiters can read directly.

  1. Entry 01

    Tech Director

    USACS

    2025 to present

    Owns

    Technical programming for Rutgers' premier undergraduate CS organization.

    Produces

    • Project-oriented workshops where students leave with working code, starter repos, implementation guides, and shippable project artifacts
    • Cross-club technical collaborations across full-stack, AI tooling, developer infrastructure, and applied CS topics
    • UHACCS 2027, the USACS flagship hackathon

    Why it matters

    Student tech events are most valuable when they end with a shipped artifact, not a slide deck.

    Proof

    UHACCS 2027 in planning; recurring workshop series in production

  2. Entry 02

    Admin, Venture Lab

    Grey Matter Society

    2025

    Owns

    Curriculum and operating structure for the GMS Venture Lab, a 4-week student-run health-tech accelerator.

    Produces

    • Weekly deliverables, team logistics, and partner outreach
    • Founder workshops, technical roadmap development, user validation, and pitch readiness
    • Final demo day for 8 teams across idea validation, MVP, and monetization strategy

    Why it matters

    Student founders need an operating cadence, not another slide template.

    Proof

    8 teams guided from idea to functional MVP and pitch readiness

  3. Entry 03

    Co-Director of Competition

    Rutgers Data Science Club

    2025 to present

    Owns

    Two Fall 2026 datathons including sponsorship, budget, prize structure, logistics, judging, workshops, and marketing.

    Produces

    • Sponsor sourcing and prize structure
    • Workshop programming for participating teams
    • Datasets, judging criteria, and post-event documentation

    Why it matters

    Datathons are a fast way for students to produce real data work that recruiters and faculty can read.

    Proof

    2 datathons in planning for Fall 2026

  4. Entry 04

    IT Intern

    Voya Financial

    Summer 2025

    Owns

    Internal automation, data migration, and Copilot-assisted automation training.

    Produces

    • Python ETL script automating large-scale data migration, reducing manual workload by 70 percent
    • Power BI decomposition tree for executives, cutting time to read data hierarchies by about 75 percent
    • Training program for 7 employees on Copilot-assisted automation
    • Contributions to company-wide AI education resources

    Why it matters

    Real internal automation is where AI tooling either compounds or stalls. I shipped the parts that compound.

    Proof

    Cross-team work with infrastructure engineers on security and operational alignment

  5. Entry 05

    Founder and Builder

    Product systems

    2024 to present

    Owns

    Rolemate, Frenzy, and VibeCheck as standalone products.

    Produces

    • Rolemate: proof-of-skill engine moving toward a Rutgers beta
    • Frenzy: multi-agent pitch evaluation platform in production hardening
    • VibeCheck: published Python CLI for AI-assisted coding guardrails

    Why it matters

    The clearest signal of capability is a system that someone outside my head uses.

    Proof

    Three live products across web, agents, and developer tooling

Capability Index

What I build with

Tools, stacks, and concepts I use in shipped systems. Grouped, not piled.

  • 01

    Languages

    • Python
    • JavaScript
    • TypeScript
    • Java
    • SQL
  • 02

    Frontend

    • React
    • Next.js
    • Tailwind
    • shadcn/ui
  • 03

    Backend

    • Node.js
    • Vercel Serverless Functions
    • Supabase
    • PostgreSQL
    • RLS
    • REST APIs
  • 04

    AI Systems

    • OpenRouter
    • LangChain
    • RAG
    • Vector Search
    • Structured Outputs
    • Tool Calling
    • Multi-Agent Swarms
    • LLM Orchestration
  • 05

    Dev Tools

    • Git and GitHub
    • Postman
    • CI/CD
    • Claude Code
  • 06

    Concepts

    • MCP
    • Agent orchestration
    • Retrieval systems
    • OAuth
    • Client-server architecture
    • Usage quotas
    • Technical education
    • Product validation
    • AI safety
    • Model evaluation
Founder Notes

Working thoughts

Notes I'm developing on proof-based systems, student recruiting, AI coding tools, and the operating model behind student technical programs.

  • Memo 01Coming soon

    Early-career recruiting is becoming proof-based.

    AI makes application text cheaper. Credible evidence becomes the scarce signal.

  • Memo 02Coming soon

    Student projects become valuable when they produce evidence people can trust.

    A project should show judgment, execution, and role-specific capability.

  • Memo 03Coming soon

    AI coding tools need understanding gates.

    The next risk is merging code people cannot explain.

  • Memo 04Coming soon

    The best student tech events end with shipped artifacts.

    Workshops should produce working code, starter repositories, implementation guides, or live demos.

  • Memo 05Coming soon

    A useful AI system turns scattered context into a decision.

    The product challenge is turning messy user material into a next action.

Reading Guide

How to read this site

Quick answers to the questions a recruiter, founder, or collaborator might ask while scanning. The throughline behind everything here is structured proof.

  • Q01

    Are you too broad?

    The throughline is structured proof. Rolemate maps student proof to jobs. Frenzy maps startup ideas to investor-grade objections. VibeCheck maps AI-generated code to demonstrated understanding. USACS, the Rutgers Data Science Club, and Grey Matter Society connect to helping technical students build, evaluate, and present real work.

  • Q02

    What kind of roles are you interested in?

    Full-stack engineering, AI systems, developer tooling, product engineering, data-intensive applications, startup engineering, and technical education.

  • Q03

    What should someone contact you for?

    AI products, student recruiting, developer tooling, multi-agent systems, Rutgers tech programming, startup and student founder projects, and technical education.

  • Q04

    What makes this different from a resume?

    A resume lists roles. This site shows the systems, inputs, outputs, and proof behind them.

Side BJazz musician

Beyond code

I'm also a jazz musician. Ten years of saxophone, ensemble performance, improvisation, and arranging shaped how I think about structure, timing, iteration, and creative systems.

Contact

Get in touch

Reach out if you're building in AI, student recruiting, developer tooling, startup infrastructure, or technical education.