I architect low-latency systems from the metal up — where microseconds and correctness both matter.
Experience
Market-Making / Liquidity-Provision (MM/LP) Systems Developer
Software Engineer Intern → Returning Software Engineer Intern
CIS 5480 — Operating Systems (graduate, ~250 students)
Global Technical Product Management Intern
Flagship Work · Korea Investment Securities
The two modernizations I lead that are moving the firm's legacy trading platform onto a faster, cleaner foundation.
A ground-up re-platforming of the firm's 13-service .NET trading monolith into independently deployable services — migrated piece by piece and verified against the old engine running in parallel, so behavior is provably preserved with no risky all-at-once cutover.
A from-scratch order/hedge engine — core-pinned threads, busy-spin loop, lock-free queues, pool allocator — that cut tick-to-trade latency from 5 ms to ~25 µs (200×); the basis of porting the core order logic from Windows/C# to C++20.
Independent Projects
Self-directed work at the intersection of markets, systems, and frontier tooling.
A local, self-hosted multi-agent system that researches and scores equities — parallel agents gather and cross-check multi-source data, a multi-factor model scores each name, and a self-audit loop of independent agents adversarially re-grades every verdict before it surfaces in the web dashboard.
An end-to-end encrypted election system where ballots are tallied entirely under encryption — so a server or DB compromise (SQL injection, exfiltration, tampering) yields only ciphertext, neutralizing whole classes of attacks. Led as senior design; won the "Most Technically Challenging" award.
GitHub →A private mesh — Mac orchestrator, always-on mini, and a wake-on-LAN Linux server — running scheduled data jobs and services with git-backed sync and a trace-free, reproducible deployment philosophy.
Earlier Research — ML & Cryptography
Where the ML and cryptography interests started.
A neural-network approach to summarizing and labeling academic papers.
GitHub →Document-AI extraction from receipt images with the Donut visual encoder-decoder model.
Master's thesis exploring OPRFs and their applications in privacy-preserving protocols.
Research & Education
Capabilities
Let's talk
Open to architecture, engineering, product, and business-development roles. Reach out — happy to walk through any of the work above.