Ameya Wagh

About Me

I am a Sr. Machine Learning Engineer and Roboticist with a seven + year track record of engineering and researching real-time autonomous systems. My core specialization lies at the intersection of CV/ML, and robotics.

My professional experience involves core development on perception and behaviors for autonomous vehicle systems, including commercial autonomous vehicles (passenger and truck fleets). My research experience include joint perception and manipulation tasks for highly sophisticated humanoid robots such as the Boston Dynamics ATLAS humanoid robot. My research experience also extends in computer vision for biomedical applications.

My research focuses primarily on tackling complex Perception Challenges by advancing methods for 3D object detection, semantic reasoning, and efficient scene reconstruction. I leverage Python, C/C++, CUDA, and modern machine learning frameworks (PyTorch, TensorFlow) to develop and deploy high-performance models targeting various embedded devices in robotics and autonomous driving applications. This research focus is backed by a strong publication history in leading academic venues.

Highlights

⭐ Core Contributor: Rivian Gen 2 Autonomy Perception Stack

I was core member of the team that developed Gen 2 Autonomy Perception Stack, a large multimodal ML model utilizing a Transformer-based backbone for early sensor fusion. This model drives the in-vehicle World Model for objects, maps, occupancy and behavioral predictions.

Watch the 2024 Rivian Investor Day Presentation (starts at 1:10:46)

⭐ Early ML Team Member: Torc Robotics Perception Stack

I was an early member of Torc's ML team contributing to the perception stack for autonomous trucking. This included lane and free space estimation, auto-calibration, dynamic occupancy grid mapping, and lidar pointcloud segmentation.

Watch Torc Robotics Autonomous Truck Demo

Professional Services

Conference Reviewer

International Conference on Computer Vision (ICCV-WiCV) 2025
International Conference on Learning Representations (ICLR-FMWild) 2025
Association for the Advancement of Artificial Intelligence (AAAI-MURE) 2025
Association for the Advancement of Artificial Intelligence (AAAI-MARW) 2024
Association for the Advancement of Artificial Intelligence (AAAI-UC) 2024
Neural Information Processing Systems (NeurIPS-lxai) 2024 – 2025
Machine Learning for Health (ML4H) 2024

Journal Reviewer

IEEE-ACCESS 2025 – present

IEEE Working Group Member

Part of the IEEE Standard Working Group P1955 developing IEEE standards for 6GRobo. (More Info) 2024 – 2027

Publications

Loading publications...

Projects

LLama2.cpp - Standalone C++ Library for LLM Inference

A standalone C++ library for llama2 LLM inference. Developed tensor and linear algebra libraries, Transformer inference framework with support for FP32 and Int8 quantization.

TOUGH - Transportable Opensource UI for Generic Humanoids

C++ software library providing API for controlling humanoid robots (Atlas, Valkyrie) powered by IHMC controllers. Includes robot control, motion planning, navigation, and perception capabilities.

A3C Agent on TORCS Simulator

A server-worker based distributed framework implementing Asynchronous Advantage Actor-Critic (A3C) algorithm for autonomous racing in the TORCS simulator.

Self-Driving Car Nanodegree Projects

Collection of 9 projects from Udacity's Self-Driving Car Engineer Nanodegree covering perception, sensor fusion, localization, and control.

ABU Robocon

International robotics competition projects and autonomous robot development.

Electronic Tabla (MIT Media Labs)

Design Innovation initiative project. Interfaced capacitive touch sensors to traditional tabla with Arduino and ABLETON for modern sound processing and filters.