Production AI & Cloud Systems
Built OpenAI/ChatGPT integrations and deployed Claude Code agents across production applications while operating containerized workloads across AWS EKS and GCP.
Kubernetes · Prometheus · Grafana · CI/CDFULL-STACK ENGINEERING · AI SYSTEMS · PRODUCT
Ideas.
Engineered.
I’m Safwat. A full-stack engineer building AI-native products from interface to infrastructure.
HOW I WORK
Understand → define → design → build → iterate.
My strongest work sits at the intersection of engineering, product thinking and systems design — where understanding the real problem matters as much as knowing how to build the solution.
01
ACRELYX AI
Co-Founder · Founding Software Engineer
An AI-native decision platform for real-estate investors.
Acrelyx is a two-person real-estate technology startup I co-founded and architected from zero to one. The product unifies property discovery, investment analysis, comparable-sales research, investor criteria and portfolio workflows into a single decision surface.
Property discovery
AI deal analysis
Buy Boxes
Saved deals
Comparable sales
Investor workflows
Full-stack architecture
Built across React.js web and React Native/Expo mobile frontends with Node.js/Express APIs, PostgreSQL/Prisma, Redis/BullMQ asynchronous processing and pgvector infrastructure for vector search and AI-driven analysis.
React.js
React Native / Expo
Node.js / Express
PostgreSQL / Prisma
Redis / BullMQ
pgvector
LLM workflows
AI as product infrastructure
Integrated LLM analysis directly into the core decision workflow, combining listing data, comparable sales, valuation signals and investor criteria to produce structured deal scores, verdicts, reasoning and recommendations rather than exposing users to a generic standalone chatbot.
Product discovery & measured impact
Worked directly with real-estate operators and an experienced business attorney to translate transaction, risk and decision workflows into product requirements. Internal benchmarking showed an 80%+ reduction in first-pass analysis time and 5× greater property-screening throughput.
Commercial source code remains private. Product screens shown here omit sensitive
implementation details.
02
PALATRA
Founder · Product Lead
Experience-first restaurant and venue discovery.
Traditional restaurant search is good at telling users where to eat, but much worse at answering what the experience will actually feel like. Palatra reframes discovery around intent, context and atmosphere.
The problem
Important dining decisions often require bouncing between menus, reviews, social media, reservation apps and restaurant sites.
The product
Users refine recommendations through vibe, cuisine, occasion, price, interior and outdoor experience before exploring curated matches.
Product strategy
UX architecture
Interaction design
Information architecture
Frontend prototype
03
SENIOR CAPSTONE · CLASSROOM ATTENDANCE TRACKER
Full-Stack Development · UI / Frontend
One attendance system, two completely different user flows.
Designed and implemented a role-based classroom attendance platform with separate teacher and student experiences inside one consistent system.
Teacher
Create class → Generate QR → Run session → Review activity
Student
Create account → Scan classroom QR → Attendance recorded
Engineering challenge
Built RESTful Flask APIs for authentication, roster processing and attendance workflows while keeping instructor administration powerful and student check-in nearly frictionless across desktop and mobile layouts.
Python
Flask
REST APIs
Role-based UX
Responsive design
QR workflows
SELECTED PRODUCTION & TECHNICAL WORK
Design taste matters. Shipping systems that move a metric matters more.
Customer Retention Platform
Built a Flask-based internal messaging platform with authenticated service-to-service communication, webhooks, retries, rate limiting, logging and reliable message processing.
40% increase in returning-buyer engagementAutomated Auction Proof Capture
Built an automated system that detected live eBay lot changes, captured and labeled video evidence and integrated the results into an existing Flask review workflow.
26% faster order-shipping velocity during pilot runsAuction Analytics & Data Pipelines
Developed Python and SQL ETL workflows and reporting around buyer behavior, geographic demand, seasonality and auction performance.
Data pipelines → operational decision supportTinyJava Compiler
Built a compiler pipeline implementing lexical analysis, LALR parsing, AST construction, symbol tables, scoping, type checking and semantic validation.
View compiler on GitHub ↗
ABOUT
Technical enough to build the system. Product-minded enough to question whether it should be built that way.
I’m a full-stack engineer with a background in computer science, mathematics, data science, distributed systems, cloud infrastructure and early-stage product development.
I’m most energized by ambiguous problems where I can work directly with users, understand the real problem, define the architecture and turn an idea into production software.