Khoi Huynh

CS student at the University of Oklahoma focused on systems programming and backend development in C/C++/C#/Java — building low-level data structures, memory allocators, and OS simulations, with a strong focus on performance and debugging. Seeking a Software Engineering internship on a systems or backend team.

Custom Memory Allocator in C | 2026

• Designed and implemented a custom heap allocator in C (my-malloc) using an intrusive free-list with boundary-tag coalescing, sbrk()-based heap growth, and direct mmap()/mremap() for large allocations above a 128KB threshold
• Added thread-safe concurrent access via a global mutex guarding all free-list mutation paths (malloc, realloc, free, split/coalesce/expand), and directed concurrent-load test generation to validate correctness under multithreaded access.
• Debugged 6+ allocator correctness bugs including a stale next-fit pointer causing a silent double-free/aliasing bug, fixed by advancing rover to the unlinked block's successor rather than resetting to head — preserving next-fit search locality instead of degrading to repeated first-fit scans.
• Directed a fork-isolated test suite (37/37 basic, 103/103 edge-case, 144027 multithreaded checks, ASan/UBSan clean) with nested-fork handling for expected-crash cases, plus a percentile-based (p50/p90/p99) benchmark harness comparing against glibc.

libds — C Data Structures Library | 2026

• Designed and implemented 4 linked-list variants in C (tagged-union, void*, macro-generic, intrusive) to empirically compare type safety, allocation cost, and dispatch overhead across common generic-programming strategies in C
• DIrected and build a common benchmark harness to test all four variants under identical workloads, measuring p50/p95/p99 latency to compare performance rather than relying on average-case timing alone
• Caught and corrected a methodology flaw where variants were being benchmarked at mismatched input scales, then re-ran and republished results once the comparison was apples-to-apples
• Validated correctness and memory safety across all four implementations with Valgrind/ASan/UBSan, using failures from generated test cases to drive debugging

Operating System Simulation | 2025 

• Built a CPU scheduling simulator in C++ modeling process lifecycle (NEW→READY→RUNNING→TERMINATED), implementing FCFS and round-robin scheduling with context-switch overhead and I/O-wait state transitions
• Implemented a variable-partition memory allocator simulating segmentation, including a free-block list with adjacent-block coalescing to reduce external fragmentation
• Designed a multithreaded parallel-sum algorithm using a binary tree of pthread workers, synchronizing compute and termination phases with mutex/condition-variable pairs across parent-child thread relationships
• Reimplemented the same parallel computation using fork()-based multiprocessing and pipe() IPC instead of threads, comparing shared-memory vs. message-passing concurrency models for the same problem