Rishav Dutta

AI Researcher

Sunnyvale, California, United States6 yrs 6 mos experience

Highly StableAI Enabled

Key Highlights

Led high-impact AI projects at SoftBank.
Developed Japan's first Handwritten OCR system.
Expert in Robotics and Machine Learning applications.

Stackforce AI infers this person is a Robotics and AI specialist with extensive experience in developing machine learning solutions.

Contact

Skills

Core Skills

Artificial Intelligence (ai)Computer VisionMathematical ModelingPython (programming Language)C++Ai System DesignOptical Character Recognition (ocr)Image ProcessingRoboticsMachine Learning

Other Skills

LinuxMonocular Depth EstimationPyTorchComputer GraphicsDeep Neural Networks (DNN)Scholarly ResearchSLAM3D Object DetectionAutonomous NavigationExtended Kalman FiltersGenerative Adversarial Networks (GANs)Model PruningGenerative Neural NetworksConvolutional Neural Networks (CNN)Reinforcement Learning

About

I am a Senior Machine Learning Applied Research Engineer at Softbank Corp., on a sabbatical to pursue Masters at University of Massachusetts Amherst. As a graduate researcher, my area of interest lies in Neural Networks, Computer Vision and Reinforcement Learning. My goal is to tie them altogether in Robotics application. My industry experience focussed majorly on research and developement of machine learning based solutions in the domain of Image Processing and Robotics. I have been responsible for delivering several high impact products building from nascent POC upto deployment phase.

Experience

6 yrs 6 mos

Total Experience

2 yrs 10 mos

Average Tenure

10 mos

Current Experience

Wayve

Machine Learning Engineer

Jul 2025 – Present · 10 mos · Sunnyvale, California, United States

Driving features research and engineering

Artificial Intelligence (AI)Computer Vision

Microsoft

Graduate Student Researcher (Extern)

Jan 2025 – Jul 2025 · 6 mos · Amherst, Massachusetts, United States · Hybrid

Developing a General Framework for an Agent-as-a-Judge System
Conduct step-by-step analysis of agentic processes.
Minimize computational overhead and reduced dependence on resource-intensive fine-tuning.
Leading to reliable evaluation across diverse domains.

Mathematical ModelingPython (Programming Language)

Softbank

3 roles

Senior Machine Learning Project Lead

Promoted

Apr 2023 – Aug 2024 · 1 yr 4 mos

I have driven high-impact AI projects, including the development of SoftReef, a demo application incorporating 3D object detection, handwriting generation, and image super-resolution.
I played a pivotal role in the Tsukuba Challenge, leading the design of robust SLAM and autonomous navigation modules by fusing GPS, IMU, LiDAR, and stereo camera data with Extended Kalman Filters. I also developed heuristic-based path planning algorithms and ensured compliance with dynamic traffic regulations for autonomous vehicles.
My leadership extends to deploying scalable cloud-based AI solutions and mentoring junior engineers, fostering innovation across AI and robotics teams.

C++LinuxAI System Design

Senior Machine Learning Engineer

Promoted

Apr 2021 – Mar 2023 · 1 yr 11 mos

I developed Softbank’s first Japanese Handwritten OCR system, achieving 91% end-to-end accuracy and enhancing downstream processes like fraud detection. To support model training, I created a synthetic handwriting text generator using conditional GANs and image processing techniques, reducing development costs by 80%. I also engineered a modular training pipeline leveraging Python’s multiprocessing capabilities, achieving a 400% speedup. My contributions extended to building a document-image orientation correction system with 1° angular resolution and compressing OCR models using Quantization and Pruning techniques, reducing model size by 66% and inference time by 70%.
Beyond OCR, I contributed to real-time monocular depth estimation for drones and vision-based inventory maintenance systems for supermarkets, showcasing my versatility in AI applications.

LinuxMonocular Depth EstimationOptical Character Recognition (OCR)Image Processing

Research And Development Engineer

Oct 2019 – Apr 2021 · 1 yr 6 mos

As an R&D Engineer at Softbank, I focused on developing deep learning models for Japanese Printed-Text OCR, achieving a 10% improvement in recognition accuracy through innovative regularization techniques. I built a Python-based text image augmentation library to address real-world issues like motion blur, reflections, and watermarks, significantly improving OCR performance on noisy data.
Additionally, I developed a ResNet-based orientation correction model for Japanese Driver’s Licenses, attaining 99% accuracy, and successfully deployed it in a production environment after rigorous commercial testing. My role required close collaboration with multiple teams, enabling the seamless integration of AI-driven solutions into various OCR-related applications.

LinuxPyTorchOptical Character Recognition (OCR)Image Processing