Soumil Chugh

Soumil Chugh

Senior AI Agent Engineer @ Kyoso

Building AI agents that understand market trends, brand identity, and creative design to help creators produce compelling on-brand content through natural language.

Learn More View Projects

About Me

Current Role

Senior AI Agent Engineer at Kyoso, building an AI agent platform for designers and creators. Working with multimodal LLMs, deep research agents, Graph RAG, and OCR to help brands generate unique ideas for images and video ads using natural language.

Core Expertise

Specialized in LLM agents, RAG systems, and multimodal AI. 4+ years building browser automation agents, deep-research systems, gesture recognition, and intelligent image editing. 8 patents in AI/ML.

Background

Previously at Huawei leading LLM agent development. MASc from University of Toronto in eye tracking and computer vision. Published at ETRA and ICPR conferences.

Experience

Senior AI Agent Engineer

Kyoso

2025 - Present

Building an AI agent platform for designers and creators using multimodal LLMs, deep research agents, and Graph RAG. Developing agents with tools and memory to understand market trends, brand data, and creative designs. Leveraging OCR for document understanding to generate unique ad ideas for images and videos through natural language.

Senior AI/ML Research Engineer

Huawei - Human Machine Interaction Lab

2020 - 2025

Led development of LLM agents for intelligent image editing, browser automation (similar to GPT-Operator), and deep-research RAG systems. Built gesture recognition with LLM function calling. Fine-tuned LLMs using LoRA/PEFT. Advanced gaze tracking with 25% accuracy improvement.

Machine Learning Consultant

General Prognostics Inc.

2020 - 2024

Developed ML pipeline for heart disease prediction using smartwatch data achieving 70% accuracy. Built computer vision algorithms for biological sample quality control. Contributed to multiple patents.

Senior Software Engineer

Jana Care

2015 - 2018

Spearheaded FDA-approved smartphone-controlled robotic system for automating blood tests. Developed BLE stacks on Android and ARM Cortex-M4 platforms.

Education

Master of Applied Science in Electrical and Computer Engineering

University of Toronto

2018 - 2020

Supervisors: Prof. Jonathan Rose and Prof. Moshe Eizenman

Research Focus: Eye tracking systems for virtual reality, computer vision, and human-computer interaction

Thesis: "An Eye Tracking System for a Virtual Reality Headset" - Developed a novel head-mounted eye tracking system achieving 1.1° gaze accuracy, 100% better than current VR systems

Expertise

LLM Agents
Browser automation, workflow agents, deep-research systems, and function calling architectures
RAG Systems
Retrieval-augmented generation, multi-source content merging, and automated visualization
LLM Fine-Tuning
LoRA, adapter tuning, PEFT techniques for domain-specific model optimization
Multimodal AI
Intelligent image editing, gesture recognition, and vision-language models
Computer Vision
Gaze tracking, pupil segmentation, deep learning for eye tracking systems
Tech Stack
Python, PyTorch, LangChain, OpenAI, TypeScript, Docker, GCP

Get In Touch

soumil.chugh@gmail.com