Aaryush Gupta

AI Engineer building compliant, scalable products that ship.

Business-minded, technically rigorous. I design and deploy AI systems end-to-end.

View Projects Download Contact Card
Madison, WI aaryush@outlook.com (949) 331-2850
Headshot of Aaryush Gupta

About

I’m an AI-first engineer and product leader. I’ve led architecture and delivery for HIPAA-compliant platforms processing millions monthly, shipped full-stack products from MVP to scale, and aligned teams around clear roadmaps. I balance rigorous engineering with pragmatic product instincts to move fast without breaking trust.

Experience

Cylerity — Lead AI Engineer (Jan 2024 – Present)

Own architecture, AI systems, and team for a fintech-healthcare startup underwriting loans on $4B+ claims data; targeted AR reduction from ~45 to 1 day.

  • Designed HIPAA-compliant infrastructure; built scalable training/inference for PHI at ~$4MM/month throughput.
  • Overhauled Azure Data Factory pipelines; improved performance and maintainability.
  • Shipped full-stack customer web app (Next.js, Node.js, Firebase, SQL Server, Azure App Service).
  • Implemented Confluence/Jira workflows; supported fundraising (high 6 figures + 8-figure lending facility).

Skale.win — Founding Engineer/Head of Product (Jun 2024 – Present)

Stealth platform that simulates interviews and gives instant, role‑specific feedback. I led product and engineering from MVP, building it to adapt quickly as AI tools change

  • Stealth GenAI for interviews and upskilling; built MVP and modular architecture.
  • Integrated ElevenLabs, STT, reasoning models; on-demand prompt tuning for rapid iteration.
  • Perplexity-powered tailoring; migrated to OpenRouter for multi-vendor flexibility.
  • ~100 beta users; GTM Q3 2025. Stack: FastAPI, Next.js, Kubernetes, Azure, OpenRouter, ElevenLabs, Perplexity.

AI Engineering Consultant — Founder (Dec 2024 – Present)

I help teams ship AI features that make a real dent—clean data in, clear value out.

  • Advisory and builds across healthcare, education, and innovation ecosystems.
  • UMN Discovery Launchpad: AI ingestion + full-stack app; automated progress tracking & gap detection.
  • Univise: RAG chatbot; 2,000 users in first week; best pitch at Transcend.
  • SmartBenAI: HIPAA-compliant claims parsing using Gemini Vision.

Selected Projects

Cylerity Claims Underwriting Platform

A platform that helps healthcare providers get paid faster by advancing payments against insurance claims.

Enabled 8-figure lending facilityMigrated MVP→scale in Azure Data FactoryPHI throughput ≈ $4MM/month
Next.jsTypeScriptAzure

Cylerity Claims Underwriting Platform

Advance cash to healthcare providers in days, not weeks.

Problem — Clinics wait weeks for claims to pay, creating cash crunches that stall operations.

What I built

  • Uploads and reviews claim data, then quickly estimates risk and payback timing.
  • Gives clinics near‑instant decisions so cash arrives in days, not weeks.
  • Dashboard shows what was approved, what needs attention, and why.
  • Built to protect patient information and meet healthcare privacy rules.
  • Helped unlock financing so the business could scale quickly.

Result — Providers receive decisions fast and see exactly what was approved and why. The business unlocked financing to scale underwriting confidently.

AR lead time
≈ 1 day
Monthly throughput
$4M+ PHI
Financing unlocked
8-figure

Skale.win GenAI Interview Platform

Practice interviews and faster candidate reviews with realistic conversations and instant feedback.

~100 beta usersGTM Q3 2025Rapid iteration via prompt tuning
FastAPIPythonNext.jsTypeScript

Skale.win GenAI Interview Platform

Practice realistic interviews and get instant, targeted feedback.

Problem — Job seekers and hiring teams waste time with generic prep and slow review cycles.

What I built

  • Speak your answers and get helpful guidance on clarity, structure, and depth.
  • Generate role‑specific questions in one click and tailor resumes to job posts.
  • Hiring teams can review answers quickly and keep everything in one place.
  • Designed to adapt as AI tools change without locking users in.

Result — Candidates get better faster with role‑specific coaching. Teams review answers quickly and keep everything in one place.

Beta users
~100
GTM target
Q3 2025
Iteration speed
On-Demand

UMN Discovery Launchpad

A dashboard for founders and mentors that turns interviews and documents into a simple plan of next steps.

Automated gap detectionTimely expert nudgesSecure multi-tenant data
Next.jsTypeScript

UMN Discovery Launchpad

Turn interviews and documents into a clear plan of next steps.

Problem — Early founders juggle scattered notes and docs; it’s hard to know what to do next.

What I built

  • Brings pitch decks, research, and mentor notes into one organized view.
  • Highlights missing information and suggests what to do next.
  • Tells mentors where their help will have the most impact.
  • Built with privacy in mind for early‑stage companies.

Result — Teams see gaps at a glance and get timely nudges when mentor help will matter most. Private data stays private.

Gap detection
Automated
Content Delivery
Real-time
Hosting
Vercel

Univise (Academic Advisor AI)

A student helper that answers common advising questions and points to the right university pages.

2,000 users in week oneBest pitch at Transcend
RAGNext.jsTypeScriptLangChain

Univise (Academic Advisor AI)

Answer advising questions in plain language with links to official pages.

Problem — Students struggle to find concrete answers across scattered university docs.

What I built

  • Finds information about courses, deadlines, and policies in plain language.
  • Grew quickly on campus thanks to word of mouth.
  • Built with RAG and LLMs to be fast and accurate.

Result — Quick answers with source links boosted adoption and saved advisors time.

Week‑1 users
2,000
Recognition
Best Pitch
Adoption
Viral on Campus

SmartBenAI Claims Adjudication Automation

Reads insurance claim documents and pulls out the important fields to reduce manual data entry.

High accuracy parsingCompliance-first design
AzureNext.js

SmartBenAI Claims Adjudication Automation

Extract key fields from claim documents to cut manual data entry.

Problem — TPAs spend hours re‑typing claim data from noisy scans.

What I built

  • Users upload PDFs or images and receive clean, structured claim data.
  • Flags low‑confidence items for a quick human check to avoid mistakes.
  • Built for healthcare standards so sensitive data stays protected.

Result — Clean, structured claim data with human checks where needed; built for healthcare privacy standards.

Accuracy
High
Compliance
HIPAA-ready
Status
Shelved

Skills

Product & Full‑Stack

Next.js TypeScript Node.js FastAPI PostgreSQL

AI/ML & Data

PyTorch LangChain RAG SQL

Cloud & Infra

Azure AWS Docker Kubernetes

Ops & Tooling

OpenRouter Perplexity HuggingFace