Hi there, I'm Ahnaf An Nafee

Bridging the gap between cutting-edge AI research and production-grade engineering.

🌐 Portfolio • 📄 Resume (PDF) • 🎓 Google Scholar • 🔬 ORCID

I'm a Computer Science PhD student at George Mason University, focused on human-computer interaction and interactive systems, with anticipated graduation in June 2029. I have worked with the DCXR Lab since January 2026 and held a Graduate Research Assistant appointment in Summer 2026. Before grad school, I was a DevOps Engineer at Mindex, working as a contractor for Paychex, and CTO at Dynasty 11 Studios.

My work combines Unity/C# and Python research tooling with mobile/backend engineering, recommendation pipelines, and reproducible performance evaluation. I am seeking Summer 2027 research internships in HCI, human-centered AI, graphics, and ML systems.

• 🔭 Current work: interactive research software, high-level classification tooling, and reproducible computational evaluation
• 🌱 Currently learning: neural fields, diffusion models, differentiable rendering
• 💬 Ask me about: AI × graphics, Kubernetes/OpenShift, CI/CD pipelines, technical leadership, game dev
• 🤝 Open to: research collaborations, open-source contributions, internships, speaking opportunities
• 📫 Reach me at: [email protected]

Selected impact: reduced cloud costs by 80% at Dynasty 11 Studios; standardized Kubernetes/OpenShift deployments across 10+ service teams at Mindex; automated TLS renewals, eliminating manual work and reported certificate-expiration downtime.

Player 2 is continuing independent work, separate from my former CTO appointment. Its recommendation pipeline uses Python/LightGBM and Java shadow scoring. Scale tests use synthetic data: a one-million-user warm-feed benchmark improved from 7.4 seconds to 0.70-1.69 seconds, while modeled mixed-workload capacity increased from 39 to 330 requests per second. These are test results, not production user counts or demonstrated recommendation-quality gains.

I also completed two independent offline recommendation studies on the public MovieLens 1M explicit-rating dataset. The temporal LambdaRank comparison did not beat its validation-selected popularity control (NDCG@10 0.108 vs. 0.190; paired difference -0.082, 95% interval [-0.093, -0.071]). A separate next-positive-rating study found a causal Transformer exceeded a GRU by 0.014 NDCG@10 (95% paired user-bootstrap interval [0.008, 0.020]), while its descriptive gap over recent popularity was only 0.0013. Both studies evaluate observed rating activity, not exposure, views, engagement, or online lift.

📄 Full details: resume.pdf

Check out my portfolio at ahnafnafee.dev/portfolio for:

• 🎮 Game development — Unity, Unreal Engine, gameplay systems, shipped titles
• 🎨 3D graphics & technical art — shaders, real-time rendering, procedural generation
• 💻 Full-stack applications — React, Node.js, TypeScript, cloud-native backends
• 🤖 Applied ML tooling: offline recommendation evaluation, shadow scoring, and local document processing

I'm always open to collaborations with researchers and industry partners pushing the boundaries of AI and graphics. Whether it's a research project, an open-source contribution, or a challenging engineering problem — reach out.

_{⭐ If you find my work interesting, consider starring a repo or following me for updates.}