Hello! 👋🏾

I’m Faiz Chaudhry, a Machine Learning engineer (since 2020) working across multimodal learning, retrieval, and geometry-aware vision. I hold a Master’s in Data Science from Tampere University (GPA 4.77/5.00). My recent work combines large-scale training and careful evaluation—running distributed experiments on LUMI, building image-text retrieval systems with metadata-aware reranking, and studying calibration/geometry where small modeling choices have large downstream effects. I’m first author of DeepBrownConrady (IEEE), which investigates camera-parameter estimation with synthetic-to-real generalization and clean ablations. I care about reproducibility (clear configs, seeded runs, versioned data) and I bring models to production with modern tooling—MLOps, AWS, and Docker—while keeping the research loop intact (telemetry, error taxonomies, regression tests).

Interests now: multimodal alignment (long-context text–vision), retrieval-augmented training signals, robust evaluation under domain shift, and geometry-aware inductive biases that transfer.

Selected highlights

-> First-author peer-reviewed paper on camera calibration via synthetic data (DeepBrownConrady, IEEE).

-> Image-text retrieval systems: embedding stores, dedup/cluster logic, metadata-conditioned reranking, user-facing diagnostics.

-> Production discipline: experiment tracking, artifact/version management, CI for eval, deployments with MLOps, AWS, Docker.

-> Large-scale experiments on LUMI (AMD GPUs): dataset curation, distributed training, reproducible pipelines.

Honors & leadership

-> Tampere University Master’s Scholarship (100% tuition); GPA 4.77/5.00.

-> Stipendium Hungaricum (fully funded, ELTE)—awarded; offer declined.

-> Co-founder, Technojawan (NGO): delivered digital-skills trainings to under-served communities.