Developing diffusion-based autoencoders and evaluation tools for 3D face reconstruction. Building modular pipelines to compare mesh topologies (BFM, FLAME, FaceVerse) and analyze structural consistency. Collaboration with CHOP on cross-topology alignment and geometric learning.

AI Engineer at Children’s Hospital of Philadelphia (CHOP)

July 1, 2025 - July 1, 2025

Developed FaceBench, a modular 3D reconstruction benchmark adopted by 4 teams and used in 10k+ mesh comparisons. Refactored evaluation pipelines, achieving 10× speedup and 100+ concurrent GPU evaluations. Designed identity-aware metric improving reconstruction accuracy by 30% over Chamfer/L2. Benchmarked 8+ SOTA models for robustness, identity preservation, and runtime efficiency.

AI Engineer at STLab (Software Technologies Lab)

September 1, 2025 - September 1, 2025

Conducted DeepGreen AI, an empirical study on energy efficiency across 6 languages and 3 DL frameworks. Ran 60+ GPU experiments, measuring 4.6× energy and 11× runtime variance between implementations. Identified optimal configurations with 2% accuracy loss and major energy savings. Co-authored paper under review contributing to sustainable AI practices.

Founder & Technical Lead at Sightcom

March 1, 2025 - March 1, 2025

Built chromosome detection system (95% accuracy, 10k+ images) reducing clinical workload by 90%. Deployed full-stack AI workflow (PyTorch, Qt/C++, ONNX) in 3 hospitals, cutting analysis time from 5 min to 15 s. Led 3 AI products from prototype to deployment, automating 90% of ML pipeline via CI/CD.