Akhilesh Warty

About

Akhilesh Warty

Machine Learning Engineer building real-time perception systems and edge AI deployment pipelines.

I build C++/ROS2 perception systems, optimize low-latency inference on NVIDIA Jetson hardware, and design cloud-based ML experimentation infrastructure.

Experience

Research Assistant — Machine Learning & Data Pipelines

Jun 2025 — Present

Boston University, School of Public HealthBoston, MA

  • Built scalable data processing and ML experimentation workflows for 100GB+ biomedical datasets, supporting large-scale model evaluation across longitudinal cohort and health datasets.
  • Developed a distributed multiverse experimentation framework evaluating 12,000+ model configurations
  • Designed a Python metadata management system standardizing 15,000+ dataset variables for consistent cohort definitions and data lineage
PythonPySpark

Graduate Student Researcher

Mar 2024 — Jun 2025

Boston University AICV LabBoston, MA

Developed computer vision pipelines for medical image analysis, combining classical techniques with deep learning models.

  • Built end-to-end pipelines integrating preprocessing, model training, and evaluation
  • Trained classification and segmentation models using transfer learning and data augmentation
  • Addressed class imbalance and evaluated performance using standard ML metrics
  • Applied classical computer vision techniques to improve image quality and feature extraction
PythonPyTorchOpenCV

Ground Station Software Engineer

Aug 2021 — Dec 2022

Space ConcordiaMontreal, QC

Developed software systems for spacecraft telemetry processing and sensor integration in a CubeSat mission.

  • Built modular systems for real-time telemetry ingestion and processing from IMU sensors
  • Developed tools for data streaming, visualization, and debugging
  • Collaborated in an Agile team on a Canadian Space Agency-supported mission
PythonC++

Skills

Python

Primary language for ML systems, data pipelines, and experimentation infrastructure — model training, evaluation, and ETL workflows.

PyTorchTensorFlowOpenCVPySparkFastAPIRayAirflow

C++

Real-time ROS2 perception nodes, TensorRT/ONNX inference, and CMake-based builds for low-latency edge deployment on NVIDIA Jetson hardware.

OpenCVTensorRTONNX RuntimeROS2CMake

Computer Vision

Object detection, tracking, pose estimation, and model evaluation — from architecture design through quantized edge deployment.

YOLORT-DETRGrounding DINOOpenCVTensorRTONNX

Mlops Cloud

Cloud ML experimentation infrastructure — spot-instance training, experiment tracking, and CI/CD for reproducible model pipelines.

DockerTerraformGitHub ActionsAWS (EC2, S3, DynamoDB)RayAirflow

Also working with

SqlAwsDockerLinux

Areas of interest

Computer vision

Object detection, tracking, pose estimation, and segmentation models.

Edge AI & robotics

Real-time ROS2 perception, TensorRT/ONNX inference, and Jetson deployment.

ML experimentation infra

Experiment tracking, dataset lineage, and reproducible training pipelines.

Real-time systems

Low-latency inference, multi-model pipelines, and strict latency budgets.

Cloud & MLOps

Spot-instance training, CI/CD, and infrastructure as code on AWS.

Data engineering

Large-scale ETL pipelines for biomedical and imaging datasets.

Connect