Billy Ntabazi

About Billy Ntabazi

AI / Deep Neural Network Engineer specializing in edge-AI, radar signal processing, and end-to-end ML solution deployment. Skilled in designing and optimizing deep neural networks and LLM-based features using PyTorch and TensorFlow, with expertise in cloud-based pipelines (AWS, Azure, GCP) and CI/CD automation. Proven ability to validate safety-critical systems, streamline model training cycles, and lead cross-team technical execution.

Experience

1. Ai Deep Neural Network Engineer

   Ambarella Inc

   Jun 2025 — Present · OH, US

   Validated end-to-end automotive radar and perception pipelines used as regression and release gates for ADAS systems, preventing high-risk regressions from reaching downstream validation and customer delivery.Executed and analyzed large-scale regression suites across signal processing and detection stages, consistently achieving 100% pre-release pass rates and improving release confidence.Performed deep root-cause analysis by tracing numerical and statistical deviations across multi-stage DSP pipelines, reducing investigation cycles from days to same-day resolution.Applied statistical detection theory to evaluate false-alarm behavior, detection stability, and noise estimation, improving robustness under challenging noise and clutter conditions.Identified and resolved edge-case failures caused by fixed hardware limits and numerical constraints, eliminating overflow-related crashes during stress testing.Verified behavioral correctness using advanced visualization and analytics (detection maps, SNR distributions, angular consistency), ensuring algorithm integrity beyond simple numeric matching.Adapted high-resolution deep learning models from PyTorch into a production inference framework using MLOps best practices, enabling early evaluation on next-generation edge AI hardware.Diagnosed and resolved severe signal-quality degradation caused by operator mismatches and quantization effects, recovering over 6 dB of lost PSNR.Partnered with compiler and systems engineers to design a custom normalization operator, restoring signal quality to >50 dB PSNR and meeting internal quality thresholds.Built automated signal-quality validation tools integrated into CI workflows, reducing manual checks by 50%+ and catching regressions earlier in development.Delivered demo-critical models ahead of sprint deadlines, directly supporting successful automotive customer demonstrations.

Education

• Maharishi International University

  Master's degree, Computer Science

• Kyambogo University

  Bachelor's degree, Telecommunications Engineering