Looking for a job

Affan Bin Usman

@affanbinusman

Message

I build scalable machine learning systems for robotic and autonomous perception.

United States

Message

What I'm looking for

I'm looking to scale deep learning systems for robotic perception, combining fundamental research, model design, distributed training, and real-time multimodal deployment.

At Gatik AI, I prototype and benchmark perception algorithms for traffic-light detection, semantic segmentation, and state estimation, helping evolve the autonomous vehicle perception stack.

I built scalable evaluation frameworks spanning object detection, tracking, and prediction, integrating Azure, Kubernetes, Airflow, Grafana, and internal visualization tools. I also improved multimodal annotation strategies through root-cause analysis of sensor time-sync anomalies, contributing to a 5–7% mAP improvement.

Previously at Honda Research Institute USA, I designed CNN-LSTM multimodal fusion architectures for a social robot processing real-time visual and audio streams. I accelerated distributed training by 30% and improved recall by 21% on imbalanced datasets.

My work spans deep learning, robotics, generative AI, reinforcement learning, and human-robot interaction. I’ve co-led a 105+ participant user study with Cornell University, published research on robot-assisted social mediation, and co-invented a patent application on relational affect modeling.

Experience

Work history, roles, and key accomplishments

Machine Learning Engineer

Gatik AI

Jul 2025 - Jun 2026 (11 months)

Prototyped and benchmarked SOTA algorithms for traffic light object detection, semantic segmentation, and state estimation using PyTorch. Architected a scalable end-to-end evaluation framework and integrated it with Azure, Kubernetes, Airflow, and Grafana.

PyTorch ROS2 Azure Kubernetes Airflow Grafana Object Detection Semantic Segmentation State Estimation Distributed Training Model Parallelism Data Caching HDF5 Real Time Inference

Research Engineer (Machine Learning)

Honda Research Institute USA

Dec 2023 - Mar 2025 (1 year 3 months)

Designed novel multi-modal fusion architectures integrating CNN-LSTM networks for a social robot. Accelerated distributed training by 30% and boosted model recall by 21% on imbalanced datasets.

PyTorch Pytorch Lightning HDF5 CNN LSTM Multimodal Fusion Distributed Training Model Parallelism Feature Engineering Dimensionality Reduction LLM Zero Shot Reasoning User Study Design Data Curation

Data Engineer (Machine Learning)

Jenius Bank

Sep 2023 - Dec 2023 (3 months)

Optimized large-scale ETL data ingestion pipelines, ensuring high-fidelity, clean datasets by reducing manual integrations by 20% for downstream analytics and ML model training.

ETL Data Pipelines Data Cleaning SQL Python

Faculty Associate

Arizona State University

Aug 2023 - Dec 2023 (4 months)

Taught engineering fundamentals to 150+ students, focusing on engineering design, research, prototyping, and coding.

Teaching Engineering design Research Prototyping Coding

Machine Learning Research Intern

Interactive Robotics Lab, ASU

May 2023 - Sep 2023 (4 months)

Developed and optimized an LSTM-based gesture recognition model, achieving 85% classification accuracy on a custom PPG sensor dataset through rigorous hyperparameter tuning.

LSTM Gesture Recognition PPG Sensor Hyperparameter Tuning Python PyTorch