Paarth Rajpal

Co-Founder

San Jose, California, United States0 mo experience

AI EnabledAI ML Practitioner

Key Highlights

Expert in AI/ML for robotics and computer vision.
Founder of innovative AI fitness platform.
Led advanced robotics projects with high precision outcomes.

Stackforce AI infers this person is a Robotics and AI specialist with a focus on Computer Vision and Machine Learning applications.

Contact

Skills

Core Skills

Computer VisionMachine Learning

Other Skills

Artificial Intelligence (AI)C++CalculusComputer vision model developmentDjangoElectrodynamicsJavaJavaScriptKerasLaboratory RoboticsLarge Language Models (LLM)Linear AlgebraModel TrainingModel quantization and fine-tuningNeural Network Architecture Design

Experience

Solo tech

Software Engineer

Jun 2025 – Aug 2025 · 2 mos · Mountain View, California, United States · On-site

Developed a novel dynamic zero-shot grasping system leveraging a Vision-Language Model (VLM), Visual Language Actions (VLA), and object detection to classify and manipulate unseen objects. The system analyzes shape, size, and texture via VLM embeddings, then selects appropriate grasping strategies using an organized VLA hierarchy. Led the end-to-end development, from concept to implementation, under strict deadlines. Authored a research paper reviewed by industry professionals and submitted for publication.

Machine LearningComputer VisionVision-Language ModelObject Detection

Team 254: the cheesy poofs

Software Engineer

Aug 2024 – Present · 1 yr 7 mos · San Jose, California, United States · On-site

Created a real-time, custom-trained object detection system which identified centimetre-accurate real-world position of targets using a single camera, allowing for easier and automated handling in robotics.
Trained the above-mentioned model from scratch, achieving a higher precision than a professionally trained model without overfitting (94% mAP vs 98.9% mAP, both at 90% + average prediction confidence). This was achieved through extensive data collection using various lighting techniques and synthetic data, created by simulating the target in high detail (with intentional random imperfections to more closely simulate real life) and getting pictures of various scenarios through manipulating the parameters of lighting, velocity, and acceleration of the target with a physics simulator.
Team lead in a project where we deconstructed and hacked a Quest 3s, repurposing it to be a viable localization/camera/vision system to be placed on the robot at a fraction of the price of what is usually used.

Computer vision model developmentSoftware DesignRobot ProgrammingMachine LearningNeural Network Architecture DesignPyTorch+2