Ayush Tiwari

Co-Founder

India5 yrs experience

AI EnabledAI ML Practitioner

Key Highlights

Founded and led AI startup delivering innovative solutions.
Expert in computer vision with diverse project experience.
Strong background in machine learning and deep learning.

Stackforce AI infers this person is a Computer Vision and AI Solutions expert with strong entrepreneurial experience.

Contact

hr.ayush@gmail.com LinkedIn

Skills

Core Skills

Computer VisionAi SolutionsMachine LearningDeep LearningEntrepreneurship

Other Skills

Artificial Neural NetworksC (Programming Language)C++Convolutional Neural Networks (CNN)Data ClassificationGANsImage ProcessingMarket ResearchMatplotlibModel OptimizationNatural Language Processing (NLP)Neural NetworksNumPyOCRObject Detection

About

Building startups, AI software and solving problems. For any freelance requirements, you can directly text me on LinkedIn. I can help you with AI, marketing, strategy and consulting. Below are things that demonstrate my technical knowledge: *Computer Vision ( semantic & instance segmentation, object detection, object tracking, GANs) *Python *Deep learning *OS: Windows, Ubuntu Reach out to me if you feel I can help you in some ways.

Experience

Clutterbot

ML

Jan 2023 – May 2025 · 2 yrs 4 mos · Bengaluru, Karnataka, India · On-site

Floor Segmentation and Mapping for Robot Navigation **
Trained and optimized a Floor Segmentation model to leverage free drivable space information for
efficient robot navigation in densely cluttered indoor environments. Utilized INT8 quantization with
TensorRT for faster inference.
Improved the model’s performance and generalization on novel classes by 17% mIoU through diffusion-
based data augmentation, knowledge distillation, and weakly supervised & curriculum learning.
Developed real-time navigation solutions based on the model’s output, including:
∗ A lightweight algorithm to generate real-time occupancy maps ingested by the Navigation Stack for
dynamic obstacle avoidance.
∗ A solution enabling the robot to traverse different floor heights (e.g., rugs) by fusing segmentation
outputs with IR sensor data.
Multi-Task Deep Learning Model for Computational Efficiency **
Designed and implemented a unified model combining segmentation and detection tasks to reduce
computational load and enhance real-time performance without compromising on accuracy

Machine LearningDeep LearningRobot Navigation

Ai2see (formerly deepvision)

Founder and Technical Lead

Mar 2021 – Present · 5 yrs

Established a computer vision services company, delivering AI-powered solutions to clients in the USA,
Europe, India and Australia.
◦ Designed and implemented an automated annotation pipeline, reducing project delivery times by 50% and cutting costs for clients during the pandemic.
◦ Delivered diverse AI solutions, including:
Provided an end-to-end solution for automated counting of bolts and nuts for a factory, managing every
step from camera and lighting setup to building a user-friendly UI for seamless client operation.
Workflow automation tools such as cylinder tracking and counting systems.
A custom automatic number plate recognition (ANPR) system optimized for Indian license plates,
deployed at scale for logistics clients.
◦ Managed and mentored a team of 4 engineers and 10 annotators, ensuring seamless delivery and quality
assurance across diverse projects.

Computer VisionAI SolutionsTeam Management

E-cell, nit delhi (entrepreneurship-cell nitd)

Founder

Sep 2020 – Aug 2023 · 2 yrs 11 mos

Started E-Cell and lead the team of 21 to cultivate entrepreneurship In college.
We started Entrepreneurship-Cell with the aim of helping budding Entrepreneurs in College.
Led the team of 21 budding entrepreneurs of NIT Delhi to build an Entrepreneurial environment in college.

EntrepreneurshipTeam Leadership

Freelance

Computer Vision Engineer

Jun 2020 – Jan 2023 · 2 yrs 7 mos

Given a target image containing a car without background, the model generates realistic shadows for the car. Experiment with different GAN architecture, finally image to image translation (Pix2Pix) GAN was used.
Segmentation of interior and exterior parts of the car. Experiment with different models, hyperparameters and augmentations to improve the results.
Chinese + English OCR for a visa, passport and business document. PANet IC15 was used for text detection and SAR was used for text recognition.
Background removal in real-time using semantic segmentation. A person is detected and localized in the image through HRNet (semantic segmentation) and then the background is
swapped with another image
Counting of different types of vehicles in a video.Yolov3 was used for detection and SORT was used for tracking.

Computer VisionGANsOCR

Fab stores

Founder

Apr 2020 – Dec 2020 · 8 mos

We did market research on 40+ Retail-Tech Startups across the World.
We try to build a platform where local retailers and consumers can transact hassle-free and delivery will be given by retailers.
Shared my research with startups like Dukaan, My Store, and Sell Nicely.

Market ResearchEntrepreneurship