Miras Sagirbay

Data Team Lead

ThanksCarbon

Leads development of a multilingual farmer-evidence platform and coordinates a team working across remote sensing, computer vision, and AI products.

• Lead the architecture and development of ASKER, a multilingual conversation-first platform using guarded LLM workflows to collect auditable field boundaries, agricultural operations, photos, and voice evidence from smallholder farmers for future digital MRV programs. • Lead the data team across remote sensing, computer vision, and AI product development, coordinating technical priorities and integration with MRV platforms.

Machine Learning Engineer

ThanksCarbon

Built satellite-imagery models for rice-field detection and delineation, forest monitoring, and photo-based irrigation classification; reported F1 78.3% and IoU 0.738.

• Developed an AWD detection model for Cambodia rice fields using Sentinel-1 and AlphaEarth; achieved a 78.3% F1 score, outperforming research benchmarks (74.2%) using a company project dataset. • Developed a rice paddy delineation model for MRV platform projects using high-resolution satellite imagery and open-source datasets, achieving an IoU score of 0.738. • Developed an AI-based forest monitoring platform using satellite imagery for forestry health assessment and deforestation detection.

Job Seeker Visa (D-10)

Active Job Search

Sought tech employment in South Korea and continued independent learning and machine learning projects after graduation.

• Actively sought professional employment in the South Korean tech industry following university graduation. • Engaged in self-directed learning and continued development of personal machine learning projects.

Research Assistant

Konkuk University

Collected Korean text corpora with web crawlers and supported research publication through editing and data visualizations.

• Developed automated web crawlers using BS4 and Selenium to collect Korean text corpora from diverse online platforms. • Supported research publication by text formatting, technical editing and creating data visualizations of research data.

Machine Learning Engineer

TengriLab

Developed speech-recognition and biometric-authentication models, reporting 10.55% WER and 96% accuracy, and managed six GPU workstations.

• Developed a telephony ASR model using Nvidia NeMo to improve user interaction and data accuracy, achieving a WER of 10.55%, an 8% improvement over the previous model. • Developed liveness detection and user recognition model, deploying them on Raspberry Pi for real-time biometric authentication, achieving 96% accuracy using a hybrid dataset of open-source and proprietary company data. • Managed 6 GPU Linux workstations, configuring remote access for AI training.

B.S., Computer Science and Engineering

Sejong University