At Outlier.ai, I evaluated and ranked multi-turn outputs from frontier large language models and speech-to-speech systems using side-by-side comparisons to support RLHF.
I analyzed text and audio outputs for factual hallucinations, safety issues, reasoning flaws, ASR/TTS defects, and prompt adherence, then wrote evidence-based rationales against project rubrics.
For the Bahnar OCR & Document Digitization Project at HCMUT, I transcribed document scans into ground-truth data and checked text normalization and spatial accuracy.
I also developed a Smart Home IoT System project with React Native, Expo Router, and WebSockets, and reproduced Elsevier’s Journal of Algebra layout in LaTeX for an academic typesetting commission.

