Headline

AI/ML Engineer & Full-Stack Developer | YOLOv8 · Docker · AWS · React | B.Tech Electronics and Computer Science (AI/ML) @ VIT Mumbai

About

I build AI systems that don't just work in Jupyter — they run in production. I'm Om Surve, a final-year B.Tech student in Electronics & Computer Science (AI/ML Honors) at VIT Mumbai, with hands-on experience building and deploying real-world machine learning systems. My work sits at the intersection of computer vision, backend engineering, and cloud deployment — where the real problems live. Here's what I've actually shipped: a microservices-based YOLOv8 object detection system deployed on AWS EC2, containerized with Docker, proxied through Nginx, and connected to a React frontend — fully production-ready. At VJTI Mumbai, I built a multimodal perception pipeline combining depth estimation (ZoeDepth), real-time detection, and a vision-language model (LLaVA-7B) — optimized for edge hardware using TensorRT. I've also architected a GenAI-powered FAANG interview prep system using FastAPI, SSE streaming, and Gemini 2.0 Flash, with a faithful Netflix-themed UI built in React + Vite. Right now, I'm finishing my final semester and actively looking for SDE or AI/ML engineering internships where I can contribute from day one. I care deeply about building things that are not just technically sound but genuinely useful — whether that's a CV pipeline on a Jetson device or a full-stack GenAI product. If you're building something at the edge of AI and software engineering, I'd love to connect. Reach me at omsurve310704@outlook.com or just hit Connect — I respond fast. #ComputerVision #MachineLearning #GenerativeAI #FullStackDevelopment #Python #React #FastAPI #Docker #AWSCloud #MLOps

Location

Mumbai Metropolitan Region

AI stack

gemini

hugging face

ollama

openai

Work experience

AI & Web Development Intern at Techligence, Jan 2026 - Apr 2026 · 4 mos, Hybrid, Built and deployed a production-grade computer vision system using microservices architecture on AWS — from model to live API. - Engineered a real-time YOLOv8 object detection service exposed via REST APIs, achieving reliable inference on constrained EC2 (t3.micro) instances through lazy loading and thread-level tuning - Containerized the full ML pipeline using Docker and Docker Compose, with Nginx reverse proxy — reducing deployment complexity and enabling consistent cross-environment portability - Diagnosed and resolved critical production issues including OOM crashes and inter-service communication failures, cutting system downtime by eliminating recurring deployment blockers - Integrated MongoDB Atlas for cloud-based data persistence and connected model inference to a React frontend, delivering a complete end-to-end ML product

Technical Team Member at Google Developer Groups VIT, Mumbai, Oct 2024 - Sep 2025 · 1 yr, Mumbai, Drove community learning and hands-on tech projects for 500+ student developers at Google Developer Groups VIT. - Led workshops and technical sessions on AI/ML and web development topics, translating complex concepts for beginner to intermediate audiences - Contributed to project initiatives connecting students with emerging technologies including GenAI and cloud tools - Coordinated event logistics and content for developer community events, maintaining consistent member engagement - Mentored junior members on development best practices and project execution as part of the core technical council

Machine Learning Intern at Veermata Jijabai Technological Institute (VJTI), May 2025 - Jul 2025 · 3 mos, Matunga East, Developed an advanced multimodal perception pipeline for real-time scene understanding, combining detection, depth, and language reasoning. - Built a spatially-aware object annotation system by fusing YOLOv8 detection with ZoeDepth and Depth Anything monocular depth estimation, improving scene context accuracy - Integrated LLaVA-7B with chain-of-thought prompting via Ollama to generate contextual, descriptive scene summaries from visual input - Optimized the full pipeline for edge deployment on Jetson Nano/Xavier using ONNX and TensorRT, achieving real-time inference within hardware memory constraints - Developed CLI tooling for multimodal perception and JSON-based scene summarization, enabling reproducible testing and modular pipeline evaluation

Web Development Intern at Suvidha Foundation (Suvidha Mahila Mandal), Apr 2025 - Jun 2025 · 3 mos, Remote, Contributed to the development of a web-based internship management system aimed at streamlining student applications and resume creation. Focused on building a responsive and user-friendly resume editing module, allowing users to input and update their details seamlessly. Key Responsibilities: Designed and implemented the resume editing section using HTML, CSS, and JavaScript to ensure a clean and intuitive user interface. Added interactivity and real-time input validation to improve user experience. Worked collaboratively with the team to integrate front-end functionality with the overall system. Participated in regular code reviews and version control using Git. Tech Stack: HTML, CSS, JavaScript, Flask, Git

Technical council core at VERVE VIT, Sep 2023 - Nov 2024 · 1 yr 3 mos, Wadala, Maharashtra, India

Education

Bachelor of Technology - BTech, Electronics and Computer Science, Vidyalankar Institute of Technology, Mumbai

Nagindas Khandwala College