Building AI-native software for real business workflows.
Computer Science, AI systems, and quantitative research.
I'm John — a 4.0 Honors College Computer Science student building production-grade AI systems and full-stack SaaS while running quantitative AI research. A prior enterprise business career is the edge behind how I build, not the headline.
Evidence over claims
Building beyond the classroom
Academic record and shipped engineering, side by side. Every number here is real.
In motion
Currently building
Active work across AI systems, research infrastructure, and full-stack engineering — happening now, not someday.
Distributed Systems
Docker, Kubernetes, deployment, and production infrastructure.
Human-in-the-Loop AI
Safe AI systems that remain under explicit human control.
Flagship · the proof
Agentic CRM
The flagship — a production-grade, AI-assisted, multi-tenant SaaS platform, designed end to end. It's the operating system of this portfolio and the evidence behind every engineering claim on the site.
A builder operating across the whole stack.
I work at the intersection of AI systems, full-stack engineering, applied mathematics, and real-world business operations. I'm early in my CS degree, but I operate far ahead of the typical first-year — designing production-grade architecture, running research, and shipping systems that hold up under real constraints, not toy projects.
Read the case study# not a toy project — real architecture$ docker compose up✓ postgres ready (multi-tenant)✓ redis ready (cache · queue)✓ api FastAPI · 120+ routes✓ worker background jobs online✓ web React + Vite on :5173auth: JWT/session · RBAC · CSRF · OAuthai: email analysis → human approval → CRM
Shipped and running inside Agentic CRM
120+ API Routes
Multi-Tenant Architecture
OAuth + Gmail Sync
AI Approval Engine
Linked CRM Execution
Dockerized Infrastructure
Kubernetes Ready
Stripe Billing
Role-Based Access Control
Human-in-the-Loop AI
Background Job Processing
Every item above is shipped and working inAgentic CRM— not a checklist of buzzwords.
Research + engineering
Two disciplines, one feedback loop
The research explores how humans and AI should make decisions together. The engineering puts those ideas into production — and measures whether they hold.
Active research
Quantitative frameworks vs AI decision systems.
Building infrastructure capable of evaluating decades of market behavior under multiple decision frameworks.
Applied research
Many of the ideas explored in the research program are already reflected in real software through Agentic CRM, where AI recommendations remain subject to human approval and measurable outcomes. Research supports the engineering; the engineering validates the research.
Engineering ecosystem
One platform, five services
The work isn't a portfolio page — it's a product platform. Here's how it's organized across johnrosca.dev.
One platform across five services — portfolio, research, demo, private app, and API.
Background
Most engineers learn the customer last. I learned them first.
I spent years working directly with businesses — from small companies to multi-billion-dollar enterprises. I didn't just sell software; I watched how organizations adopt it, where workflows break, and how decisions actually get made. That's a perspective most engineering students don't have, and it shapes every system I design.
Years across businesses — local operators to multi-billion-dollar enterprises — taught me:
- Software adoption
- Workflow friction
- Operational bottlenecks
- Customer behavior
- Organizational decision making
Let's build something serious.
I'm looking for AI-native, high-performance engineering environments where business judgment and engineering execution both matter.