Nader Alfares

Systems & ML Engineer

Ph.D. Candidate @ Penn State

I build intelligent, scalable systems at the intersection of machine learning, distributed systems, and cloud computing. My research and work span 3D data understanding, edge-cloud architectures, and full-stack applications.

Experience

Where I've worked

Graduate Research & Teaching Assistant · Penn State · Secure AI Lab

Aug 2021 – Present

Designed a multi-layer edge-cloud ML inference framework for VR/AR workloads; published at Performance Evaluation 2025 and IEEE VR 2025.
Engineered unsupervised/semi-supervised clustering pipelines for 3D point clouds and meshes; lead multimodal LLM and code-generation LLM security research.
Teach native iOS (Swift/SwiftUI) and web development plus cloud engineering labs (AWS, Docker, Kubernetes) to 35+ students per semester.

Co-Founder & Software Developer · Nuntii LLC

Sep 2020 – Present

Shipped production iOS apps (Swift/SwiftUI) and Next.js dashboards for e-commerce and restaurant clients.
Built serverless AWS backends (Lambda, DynamoDB, S3) with Stripe payments, Cognito auth, real-time order tracking, and multi-tenant CI/CD.

Computer Scientist · Anomalee Inc

Jul 2022 – Aug 2022

Built a full-stack inference console and RESTful API with Flask and PyTorch for security anomaly detection, model evaluation, and data annotation.

Research & Teaching Assistant · Penn State · Computer Systems Lab

Jan 2018 – May 2020

Built geo-distributed KV storage systems (C++) and serverless/Spark hybrid workloads optimized for public-cloud cost; contributed to VLDB '22 and Middleware '20.
Developed Kubernetes microservice deployments and containerized services; led CS teaching labs in Python, Java, and MATLAB.

Publications

Research

Peer-reviewed work on edge-cloud systems, geo-distributed storage, caching, and ML for 3D data.

Contrastive and Angle based Deep Clustering of Point Clouds

N Alfares, G Yang, D Miller, G Kesidis

Deep LearningClustering3D Point Clouds

NSF 2025

Virtual Caching with Apportioned Objects for Mobile Virtual Reality

N Alfares, G Kesidis

Edge CachingVirtual RealityMobile Systems

Performance Evaluation 2025

Virtual Reality Benchmark for Edge Caching Systems

N Alfares, G Kesidis

VR BenchmarkEdge ComputingCaching

IEEE VR 2025 (Poster)

Online VM Service Selection with Spot Cores for Dynamic Workloads

N Alfares, G Kesidis, B Urgaonkar, AF Baarzi, A Jain

Cloud ComputingSpot InstancesOptimization

IEEE Cloud Summit 2024

Container Sizing for Microservices with Dynamic Workloads by Online Optimization

N Alfares, G Kesidis

MicroservicesContainersOnline Optimization

WoC 2023

Projects

Selected work

Systems I have built for research and production, from distributed storage to XR benchmarking.

XRFlux

Benchmarking tool for edge-cloud XR systems: measures end-to-end latency, frame timing, and caching behavior for VR/AR workloads under realistic network conditions. Basis for the IEEE VR 2025 poster on edge caching for virtual reality.

PythonUnityVR/ARPerformance Analysis

Research Tool

LEGOStore

Linearizable geo-distributed key-value store that combines replication and erasure coding to cut public-cloud storage cost while preserving strong consistency. Built over the mcrouter library with memcached as cache and MySQL as backend. Published at VLDB 2022.

C++MemcachedMySQLErasure Coding

Distributed Systems

OSA — Online Simulated Annealing

Cluster resource manager that uses online simulated annealing to right-size containers and select VM services for dynamic workloads in real time. Underpins the WoC 2023 and IEEE Cloud Summit 2024 papers on container sizing and spot-core VM selection.

PythonMachine LearningOptimizationCloud

Cloud Computing

MCD-OS

Modified memcached that supports virtual caches for apportioned objects, adding object-sharing accountability across proxies to improve cache efficiency. Evaluation published in Performance Evaluation 2025.

C++MemcachedCache OptimizationDistributed Systems

Caching System

Neural-Agent Dogfighting Flight Simulator

Flight simulator where two neural agents learn aerial combat through competitive deep learning, exposing emergent dogfighting behavior between agent-controlled aircraft.

PythonDeep LearningGame DevelopmentNeural Networks

AI/ML

Technical Profile

Skills & Certifications

Domain expertise across machine learning, cloud infrastructure, systems programming, and mobile product development — with credentials nested where they apply.