class SystemsEngineer:
    def __init__(self):
        self.languages = ["C++", "Python", "Rust", "SQL", "Metal (MSL)", "CUDA"]
        self.systems = ["Buffer Pools", "B+ Trees", "Concurrency Control", "GPU Kernels"]
        self.inference = ["vLLM", "llama.cpp / GGML", "Speculative Decoding", "FP8/AWQ Quantization"]
        self.infrastructure = ["FastAPI", "AsyncIO", "Docker", "AWS", "Nginx", "Linux"]

        self.current_work = {
            "role": "AI Engineer @ WAI Global",
            "focus": "Designing a self-hosted LLM inference stack (vLLM, continuous batching, prefix caching)"
        }

        self.previous_work = {
            "role": "AI & Systems Engineer @ SmatBot",
            "impact": "Re-architected inference pipeline: -25% mean / -35% median latency"
        }

        self.open_source = {
            "vLLM": "Debugged illegal-memory-access crash in MTP speculative decoding (Issue #37035)",
            "llama.cpp": "Metal GPU backend: GGML_OP_CONV_3D (F16/F32), ported to CUDA (PR #19927)",
            "Docling (IBM)": "Re-engineered Excel table detection with Flood Fill (BFS) (PR #2778)",
            "LanceDB": "Sync/Async API parity in Python SDK via fast_search (PR #2962)"
        }

    def deep_tech_projects(self):
        return [
            "CMU BusTub: Buffer Pool Manager (ARC), concurrent B+ Tree (latch crabbing), query executors (C++)",
            "MSL kernels from scratch: memory layout, threadgroup sizing, CPU-reference validation",
            "Self-hosted LLM inference stack design: vLLM + provider-agnostic gateway + eval-gated model choice"
        ]

    def currently_exploring(self):
        return "LLM serving internals, GPU kernels (Metal & CUDA), and database internals."