|
▸ Vulkan · DX12 · Metal cross-platform C++ engine on SDL3 / SDL_GPU dispatching natively to Vulkan, DirectX 12, and Metal. Authoritative netcode with client prediction + lag compensation. PBR, cascaded shadow maps, GTAO, SSR, volumetric lighting, SMAA T2x. |
▸ 85 % of cuBLAS SGEMM on RTX 4090 Dense GEMM with shared-memory tiling, warp-shuffle reductions, and register micro-tiling. Nsight + roofline analysis on RTX 3090 / 4090. |
|
▸ Self-hosted Kubernetes, 6 GPUs, multi-site Bare-metal Kubernetes across multiple sites for vLLM serving + LoRA finetuning experiments. Prometheus / Grafana telemetry. |
▸ Kyber-1024 + X3DH + Double Ratchet E2E encrypted chat with identifier-free relays + custom anti-DPI TLS obfuscation. Go backend, TypeScript / React / Electron client. |
▸ Upgrading fintech infrastructure at Polity Network — took a single-EC2 monolith to a real AWS EKS cluster with OpenTofu + ArgoCD. Ephemeral per-PR review environments cut the release cycle ~40 %.
▸ Bringing Google Pixel SoCs up over PXE for a 20,000-phone datacenter at UCSD × Google. Living in the bootloader / kernel layer — UART debug, porting Android kernel modules into mainline Linux.
| LANGUAGES |
|
| GPU / HPC |
|
| AI / ML |
|
| INFRA |
|
| SECURITY |
|
Currently open to GPU systems, ML infrastructure, and high-performance engineering roles for new-grad 2026.
▸ Email · dkatulevskiy@gmail.com
▸ GitHub · github.com/Katulevskiy
▸ Resume · Daniil Katulevskiy CV.pdf →



