I am a dedicated AI researcher with over 5 years of professional experience, primarily focused on machine vision, computer vision, NLP and speech recognition. My expertise lies in developing cutting-edge solutions for industrial monitoring, image processing, and object detection and Speech recongion pipelines while also delving into innovative research in large language models (LLMs) and LLM agents.
I am deeply passionate about LLM development and research, with a keen interest in creating intelligent systems for document retrieval, conversational AI, and agentic designs. My hands-on experience spans computer vision, speech recognition, NLP, and generative AI, where I have implemented advanced frameworks and methodologies to solve complex challenges. I thrive in fast-paced, research-intensive environments, entrepreneurship spirit and am eager to contribute to start-up and impactful projects in AI technology.
Experiential Life Center | Sep 2024 – Present
- Built large language model systems for user engagement and operations.
- Designed RAG (Retrieval-Augmented Generation) pipelines for question-answering and recommendation systems.
- Developed patient referral matchmaking systems using AI-driven agentic approaches.
Luxi.Ai | Feb 2024 – Present
- Developed computer vision models for object detection, segmentation, classification, and image retrieval.
- Developed image-retrieval pipelines for industrial and commercial use cases.
AgentCo | Jan 2024 – Aug 2024
- Researched agent-based LLMs for advanced planning and task management solutions.
AI Land | Jul 2022 – Nov 2023
- Founded a company providing AI services and products, such as Bread Counter, an object detection-based algorithm for monitoring bakery production.
- Developed scalable solutions and led innovative projects in computer vision.
Virasad | Sep 2021 – Aug 2022
- Led a team providing machine vision services in industrial monitoring, optimization, and image processing.
- Specialized in classic feature extraction methods (SIFT, BRISK, ORB) and DNN-based algorithms, including YOLO, Mask R-CNN, and U-Net for object detection, segmentation, and anomaly detection.
- Implemented Agile and Scrum methodologies to enhance collaboration and deliverables.
SYMO | Dec 2019 – Jan 2022
- Conducted research and development in computer vision and generative models, focusing on classification and object detection tasks using CNNs and GANs.
- Built and fine-tuned models for company-specific datasets, including web crawling and data cleaning.
- Contributed to model deployment and performance optimization for various applications.
- Developed a toolkit that demonstrates how to use different language models for question-answering (QA) and document retrieval tasks using Langchain. The script utilizes various language models, including OpenAI's GPT and Ollama open-source LLM models, to provide answers to user queries based on the provided documents.
- Built an AI-powered multi-agent system for generating personalized 7-day diet and workout plans based on user inputs like age, gender, and lifestyle habits.
- Utilized LangChain, PhiData, and Weaviate for a dynamic agentic RAG design.
- Developed a Persian automatic speech recognition system trained on over 300 hours of data.
- Designed a custom architecture achieving state-of-the-art performance in Persian ASR tasks.
- Created a conversational agent capable of simulating human-like dialogue behaviors for virtual assistants and chatbots.
- Integrated use cases for customer service and interactive conversational AI systems.
- Built a license plate recognition model with PyTorch Lightning, including dataset preparation and model training.
- Contributed in a toolkit to simplify deep learning workflows and reduce coding overhead for frequent tasks.
- Achieved 99% accuracy in COVID detection by training classifiers using active learning on X-ray datasets in PyTorch and TensorFlow.
- Developed an algorithm for real-time dough tracking in bakeries using YOLO-based object detection.
Project Website | GitHub Repository
- Designed a paraphrasing tool with multiple grammar and style modes for NLP applications.
- Trained emotion recognition models using transformers and ResNet backbones, overcoming challenges of imbalanced datasets.
- Programming & Tools: Python, PyTorch, LangChain, LangGraph, Transformers, Phidata, TensorFlow, OpenCV
- Deep Learning: Generative AI, LLMs, CNNs, Transformers, GANs, Agents, Diffusion Models
- Fields of Expertise: Computer Vision, NLP, LLM-Agents, Speech Recognition, Machine Vision, Data Science
- Project Management: Agile, Scrum, Team Leadership
- Email: vkhalokhi@gmail.com
- LinkedIn: https://www.linkedin.com/in/vargha-khallokhi
- GitHub: https://github.com/Vargha-Kh
- Phone: +989396463632