Ha Le

@ha_le_dev

Full Stack Developer for React Native, Node and Python

Vereinigte Staaten
Englisch
Einige Informationen werden in englischer Sprache angezeigt.
Über mich
Full stack developer shipping React Native, Next.js, Node.js and Python in production every week. I build mobile apps, web apps, data pipelines and LLM features. I also build LLM pipelines as a contract engineer, including a model router that cut the cost of a code-conversion job by more than 13x. You get clear written updates, small reviewable changes and code you can maintain after I hand it over.... Mehr lesen

Kompetenzen

h
ha_le_dev
Ha Le
offline • 
Durchschnittliche Antwortzeit: 15 Stunden

Meine Dienstleistungen

Suchen & Bewerben
I will apply to jobs for you on real company career pages and track each one
Online Programmieren lernen
I will mentor you on your software project with code review and mock interviews

Portfolio

Arbeitserfahrung

Redwood_EDA

Software Engineer (Contract)

Redwood EDA • Freiberufler

Jun 2026 - Present • 5 mos

Contract software engineer for Redwood EDA, an open source hardware design tooling company. I built Conversion Atlas, the web console for an LLM driven code conversion pipeline, with a steppable run timeline, inline and side by side diffs, a summary dashboard and per attempt model exchanges. I designed a per step model router that replaced a single model agent loop with cheap first API calls and automatic escalation on verification failure, cutting the cost of a full module conversion from $8.22 to $0.60 at equal verification coverage. I benchmarked 10 models across six providers on identical conversion tasks and contributed more than 15 merged pull requests across two open source repositories.

University_of Central Florida

GPU Systems Researcher

University of Central Florida • Teilzeit

Feb 2026 - Present • 9 mos

Research assistant in the Unary Lab at UCF under Prof. Di Wu, working with AMD on GPU systems and power and thermal management. I built the NVIDIA backend for Chopper, a multi vendor GPU profiling framework developed with an AMD performance engineer, including vendor auto detection, NVML telemetry collection and a CUPTI kernel tracer in C++. I delivered a unified CPU and GPU timeline on a single clock domain across both vendors and implemented unprivileged CPU hardware counter collection. I also characterized Lit Silicon power reallocation policies on MI325X GPUs at 700W and 1000W using torchtitan traces.