Kathan Vakharia

Engineer @ Qualcomm

About

Kathan Vakharia is a Machine Learning Engineer specializing in production VLM/LLM systems with strong full-stack development expertise in Python, C++, and Golang. Proven track record deploying scalable AI solutions that reduced operational costs by 50% and processing times from 7+ days to 2 hours. Experienced in end-to-end ML pipelines—from custom dataset creation and model fine-tuning to serverless cloud deployment using Docker and modern MLOps practices. Holds an MTech in Computer Science from the National Institute of Technology Karnataka, Surathkal, complemented by a successful track record as a competitive programming instructor (on YouTube), demonstrating exceptional analytical problem-solving capabilities. Key Achievements - Full-Stack LLM Development: Architected complete LLM-powered inventory management platform, reducing critical processing time from 7+ days to under 2 hours through async backend engineering and production deployment on GCP - End-to-End VLM Systems: Fine-tuned and deployed production VLMs (Qwen2.5-VL) processing 2000+ documents daily with 95%+ accuracy, managing entire pipeline from custom dataset creation to serverless deployment on RunPod - Competitive Programming Education: Collaborated with ICPC World Finalist, Priyansh Agarwal to create educational content for 50k+ YouTube subscribers, developing structured learning pathways that democratize competitive programming for beginners

Country

India

City

Bengaluru

Industry

Information Technology & Services

Skill

C (Programming Language), MLOps, Generative AI Development, Large Language Models (LLM), Problem Solving, Competitive Programming, C++, Python (Programming Language), Go (Programming Language), Educational Leadership, Operating Systems, Cloud Services, API Development, Machine Learning, Deep Learning, Leadership

Experience

Engineer

Qualcomm

2026-2 - Present · 8 mos

Bengaluru, Karnataka, India

Competitive Programming Educator

TLE Eliminators

2025-6 - Present · 1 yr 4 mos

- Developing first principle based solutions for the CSES problemset.

Data Scientist

Petpooja

2024-6 - 2026-2 · 1 yr 9 mos

Ahmedabad, Gujarat, India

- Engineered a full-stack LLM-powered system that reduced the processing time of business processes from multiple Days down to an hour, featuring asynchronous APIs and scalable cloud architecture that achieved 80% workflow efficiency gains while maintaining robust error handling and logging capabilities. - Demonstrated expertise in adapting foundation vision language models to specialized domains through parameter-efficient fine-tuning, creating custom evaluation pipelines, and implementing automated data collection strategies that achieved >90% error reduction while preserving core model capabilities across multiple tasks.

Python Course Instructor

National Institute of Technology Karnataka

2024-1 - 2024-3 · 3 mos

Mangaluru, Karnataka, India

As an MTech student, I collaborated with a peer to serve as a Python Programming Instructor for CS101. We enhanced the existing course curriculum and delivered engaging lessons to a class of 80 BTech undergraduate students. My responsibilities spanned lesson planning, interactive instruction, exam development and grading, as well as individualized student mentorship throughout the course. This experience allowed me to enhance my Python expertise, teaching skills, and ability to design effective assessments.