Hey there,_
I'm a web developer and AI researcher with experience building scalable products and working with early-stage startups.
About Me
Hey, I'm Aadvait! I was born in India, raised in Dubai, and I'm now studying Computer Science and Business at Indiana University, Bloomington. I specialize in full-stack web development using TypeScript, Next.js, tRPC, Prisma, Supabase etc. I've built membership systems, internal dashboards, and end-to-end products integrating authentication, payments, and data pipelines. I've also traveled across the U.S. for hackathons, pitch competitions, and startup conferences, where I learned how to prototype fast, validate ideas, and communicate technical work to both users and investors.
My main interest lies in the mechanics of building and scaling software businesses: how developer tools and SaaS startups acquire their first customers, close B2B contracts, price their product, and measure performance through metrics like CAC, LTV, retention, and churn. I use YC, ODF, Z-Fellows companies as case studies to learn about early-stage growth strategy, like equity allocation, fundraising structure, onboarding optimization, and how cold outreach and distribution drive adoption.
Coming to the research side, I work in the Data Science and Artificial Intelligence Lab (DSAIL) at the Kelley School of Business, collaborating with Indiana University's Open Source Ecosystems (OSE) team on the NSF SAFE-OSE project — "Implementing AI-Enabled Vulnerability Management Practices to Enhance Safety and Security of Open Source Cloud Computing Ecosystems." My project contributes to benchmarking vulnerabilities produced by LLMs across IDEs, programming languages, and prompt variations, mapping CWE patterns to understand the security risks of AI-generated code.
Outside of research and startups, I lead Kappa Theta Pi at IU, the largest professional technology fraternity in the U.S., with 70+ members at IU. I'm also a Luddy Direct Admit, Hutton Honors Scholar, and Global Engagement Scholar supported by over $52,000 in scholarships. When I'm not building or researching, I'm probably playing the guitar, copying a Bob Ross painting, or at a McDonald's.
Work
Oracle
Software Development Intern - Oracle Cloud Infrastructure (OCI) Team
- ・Built cloud management portal with Next.js, TypeScript, and Oracle APEX, implementing real-time monitoring dashboards for clients like Emirates NBD and Etisalat.
- ・Enhanced microservices reliability by implementing RabbitMQ message queuing, supporting 99.9% uptime SLA for Oracle Cloud's UAE deployment.
- ・Integrated OAuth 2.0 with Oracle IDCS APIs and NextAuth.js v4.22, implementing RBAC for multi-tenant applications serving Dubai government entities.
- ・Developed Stripe payment gateway integrations for Oracle NetSuite clients with webhooks and fraud detection, ensuring compliance for e-commerce implementations.
- ・Automated infrastructure provisioning using Terraform and OCI CLI for dev/staging environments, accelerating the team's infrastructure-as-code migration.
IBM
Data & AI Intern - Watson & Cloud Pak Team
- ・Enhanced recommendation engine using PyTorch and collaborative filtering for IBM Watson pilot with MAF (Carrefour), processing 80K+ loyalty user transactions.
- ・Built ETL pipelines with Apache Spark, Pandas to process 3TB+ retail datasets, reducing preprocessing time by 30% through data optimization and feature extraction.
- ・Conducted model evaluation using scikit-learn metrics (precision, recall, ROC-AUC), contributing to 12% accuracy improvement over baseline rule-based system.
- ・Fine-tuned BERT models with Transformers for Arabic-English sentiment analysis on telecom customer tickets, achieving 93% production classification accuracy.
- ・Deployed ML models via Flask REST APIs on IBM Cloud Foundry with Redis caching, creating documentation for internal teams to integrate inference endpoints.
Research
Benchmarking LLM-Generated Code Security via Static Analysis and CWE Scanning
As part of IU's Data Science and Artificial Intelligence Lab and the Safe Open Source Ecosystems initiative funded by the National Science Foundation, this project investigates how LLMs introduce security vulnerabilities into AI-generated code. The study benchmarks insecure coding patterns across programming languages, IDEs, and prompt types using static analysis tools such as Semgrep, CodeQL, and Bandit, mapped to CWE/CVSS/OWASP standards. The goal is to establish a quantitative baseline for LLM-generated code security and identify the behavioral and architectural factors that propagate vulnerabilities across modern development environments.
Technologies
Key Achievements
- •Contributing to a multi-institutional NSF Safe-OSE project (> $30M federal initiative) advancing secure AI-assisted software development
- •Designed and implemented a benchmarking pipeline to analyze > 10,000 code samples across models, IDEs, and regions
- •Created a CWE-mapped vulnerability matrix quantifying severity, recurrence, and root causes in LLM-generated code
- •Developing comparative evaluations across U.S., EU, and Chinese model providers to study dataset bias and security drift
- •Supporting DSAIL's collaboration with IU's Open Source Ecosystems team to produce reproducible, open-access benchmarks for the research community
Using Transformers and DL with Stance Detection to Forecast Crypto Price Movement
Developed a stance-detection-based forecasting system that predicted Bitcoin price movements by combining transformer-based NLP and deep learning time-series models. Built a RoBERTa model to detect bullish vs. bearish market stances from cryptocurrency-related tweets and linked that sentiment to real-world price data through a Recurrent Neural Network. The pipeline achieved 80% stance-detection accuracy and a mean absolute error of $1,144 on price prediction, demonstrating the feasibility of using social media trends for crypto forecasting.
Technologies
Key Achievements
- •Fine-tuned a transformer (RoBERTa) for stance detection on 2.5K hand-labeled tweets, achieving 80% accuracy
- •Designed and trained an RNN to map stance data to Bitcoin closing prices with $1.1K MAE
- •Integrated transformer outputs with time-series forecasting to bridge NLP and financial prediction
- •Featured in SSRN Prebit Journal and Connected Papers, where it ranked as a central node in the “AI for Financial Forecasting” cluster
- •Cited by the University of Southampton (UK) and referenced in follow-up work on financial sentiment modeling
Experience
Kappa Theta Pi — Alpha Eta
Founder & President
- ・Founded and scaled the Indiana University chapter of Kappa Theta Pi, the nation's premier professional technology fraternity, leading 14 executive board members and 73 active members across technical and business disciplines
- ・Built an integrated recruitment system using Airtable, n8n, and Zapier automations, processing 200+ applicants per semester and centralizing evaluation, attendance, and decision workflows
- ・Oversaw all executive operations from governance and bylaws to brand design, mentorship systems, and long-term strategic planning for future pledge classes. Includes internal operations such as rush, budgeting, marketing, sponsorships, curriculum etc
- ・Onboarded Humana Insurance as a Silver Corporate Partner, establishing recurring sponsorship and professional development opportunities for members
- ・Organized cross-campus case competitions, startup panels, and technical workshops with Luddy and Kelley organizations, positioning KTP as a central tech-leadership hub on campus
Luddy Consulting Association
Project Manager: Solutions Engineering Team
- ・Managed a 10-member technical team through the planning and development of a custom, local LLaMA chatbot and client-facing support platform
- ・Designed and prototyped the frontend experience using Framer, focusing on clean UI and intuitive interaction flows
- ・Collaborated with developers to integrate the platform with Zoho CRM via API, enabling lead tracking and user data management
- ・Coordinated task delegation, timelines, and weekly standups to ensure consistent progress across frontend, backend, and automation subteams
- ・Worked closely with the client (TrueLeap) to align on feature requirements and deliver iterative updates throughout the development cycle
Projects
Locus Project
Key Achievements
- •Built in Unity using C#, AR Foundation, and ARKit/ARCore for real-time indoor navigation with <1m accuracy
- •Designed a 3D campus model with QR-based tracking and mapped over 200+ classrooms to unique navigation targets
- •Developed dynamic UI, path visualization with line rendering, and latency-optimized interactions
- •Deployed custom builds for multiple schools, adapting the architecture for different facilities
- •Featured in Khaleej Times, Gulf News, and The National for innovation in educational tech
Technologies
KTP Indiana Website & Member Portal
Key Achievements
- •Replaced the Framer prototype with a scalable production website using Next.js 15, TypeScript, and Tailwind CSS
- •Built a full Clerk + Supabase role-based access control system with 5 hierarchical user types
- •Developed a real-time internships board syncing 500+ listings/day from GitHub with cron jobs, caching, and Markdown parsing
- •Created a full member portal: announcements, GCal sync, merch store, alumni directory, and dues payment system
- •Designed a responsive UI using Radix UI primitives, infinite scroll carousels, and custom animation logic for smooth UX
Technologies
MNIST & CIFAR Real-Time Image Classification Systems
Key Achievements
- •Implemented CNNs with dropout and regularization to optimize classification accuracy
- •Built OpenCV-based real-time interfaces for live object and digit recognition via webcam
- •Achieved 91% MNIST and 90.3% CIFAR-10 accuracy using TensorFlow and Keras pipelines
- •Integrated data augmentation, Gaussian blurring, and thresholding to enhance model generalization
- •Visualized performance using confusion matrices, ROC and precision-recall curves, and confidence plots
Technologies
ThirdLayer - Semantic Google Drive Search
Key Achievements
- •Built a full-stack Next.js application with Google Drive OAuth integration and real-time file indexing across 1000+ files
- •Implemented semantic search using Transformers.js and Xenova embeddings with cosine similarity ranking for intelligent content discovery
- •Developed a multi-format content extraction pipeline supporting PDFs, Word docs, spreadsheets, and presentations using pdf-parse and mammoth
- •Created a ChatGPT-style conversational interface with streaming responses powered by OpenAI GPT-4, combining semantic search with LLM reasoning
- •Designed an intuitive folder browser with recursive Drive navigation, real-time progress tracking, and batch indexing capabilities
- •Built vector search service with in-memory storage, metadata filtering, and BM25 hybrid ranking for production-grade retrieval
Technologies
Nodebase
Key Achievements
- •Built an interactive visual workflow builder using React Flow, featuring drag-and-drop nodes, zooming, keyboard shortcuts, and live data previews
- •Integrated multiple trigger nodes (Webhook, Stripe, Google Form, Manual) and AI providers (OpenAI, Claude, Gemini) with pluggable APIs for easy extension
- •Implemented background job execution with Inngest, handling async workflows and queued actions with real-time monitoring
- •Added Better Auth authentication, Polar subscriptions & paywalls, and a full SaaS business layer with RBAC and billing management
- •Used Prisma ORM with Neon Postgres for scalable data persistence and schema migrations
- •Integrated Sentry for error tracking and AI-powered anomaly detection, plus CodeRabbit for automated PR reviews and workflow audits
- •Styled with Tailwind CSS and Radix UI, supporting dark mode, responsive layout, and theming built on Next.js 15's App Router architecture
Technologies
aadvait.com Portfolio Website
Key Achievements
- •Designed and developed a fully responsive, mobile-first portfolio with custom animations and scroll-triggered interactions using Framer Motion
- •Implemented a modular component architecture with reusable cards for Work, Research, Experience, and Projects sections, ensuring maintainability and scalability
- •Integrated dynamic routing and SEO optimization with Next.js metadata API, Open Graph tags, and structured data for enhanced discoverability
- •Built custom UI components including an animated tech stack marquee, sticky navigation with scroll indicators, and a full-screen hamburger menu with smooth transitions
- •Configured a serverless contact form with Formspark integration, custom success notifications, and client-side validation for seamless user engagement
- •Deployed on Vercel with automatic CI/CD, environment-based configurations, and optimized build performance for sub-second page loads
Technologies
Get In Touch
Contact Information
Feel free to reach out to me for collaboration, opportunities, or just to say hello. I'm always open to discussing new projects and ideas.
Phone
+1 (812) 947-8569Schedule a Meeting
Book a time on my Calendly