(May 2024 - July 2024)
Backend Developer Intern
Infopercept Consulting
constcoder={name:'Harsh Panchal',skills:['Python(FastAPI)', 'Node.js', 'PyMongo', 'React.js', 'MongoDB(NoSQL)', 'MySql', 'GitHub', 'Docker', 'AWS'],hardWorker:true,quickLearner:true,problemSolver:true,hireable:function() {return(this.hardWorker&&this.problemSolver&&this.skills.length>=5);};};ABOUT ME
Who I am?
My name is Harsh Panchal. I am a professional and enthusiastic programmer in my daily life. I am a quick learner with a self-learning attitude. I love to learn and explore new technologies and am passionate about problem-solving. I love almost all the stacks of web application development and love to make the web more open to the world. My core skill is based on Python and I love to do most of the things using Python. I am available for any kind of job opportunity that suits my skills and interests.
Experiences
Skills
Python
Java
Javascript
React
MongoDB
MySQL
PostgreSQL
Git
AWS
Docker
Firebase
C++
Tensorflow
Selenium
FastAPI
Python
Java
Javascript
React
MongoDB
MySQL
PostgreSQL
Git
AWS
Docker
Firebase
C++
Tensorflow
Selenium
FastAPI
PROJECTS
SevenSpaceRealEstate
constproject={name:'SevenSpaceRealEstate',tools: ['FastAPI (Python)', 'MongoDB', 'Cloudinary(CDN & Optimization)', 'Docker', 'React.js + Vite],myRole:Developer,Description: Developed a high-performance real estate platform enabling users to browse, search, and view property listings in real time. The backend was built with FastAPI for rapid API response, integrated with MongoDB for scalable data management, and Cloudinary for optimized image delivery. Implemented a real-time command palette (Win+K) to enhance navigation and user experience. Leveraged React Query for efficient state management, significantly reducing server and database load. The application was containerized using Docker to ensure consistent deployment and maintain 99.9% uptime across environments. Performance was a key focus, with API response time reduced to 300ms and page load time improved by 75% through backend optimization and image handling strategies.,};Person Retrieval System on NVIDIA Jetson Orin AGX
constproject={name:'Person Retrieval System on NVIDIA Jetson Orin AGX',tools: ['CLIP Model', 'ViT-B/32(Image encoder)', 'DistilBERT(Text Encoder)', 'ONNX', 'TensorRT', 'Python],myRole:Model Deployment,Demo:N/A,Description: I developed an end-to-end Person Retrieval System deployed on the NVIDIA Jetson Orin AGX edge device, designed to retrieve relevant person images based on natural language queries with minimal resource usage. The system is built on a CLIP-based architecture, utilizing ViT-B/32 as the image encoder and DistilBERT for text encoding. To ensure the model was efficient enough for real-time edge inference, we applied Low-Rank Adaptation (LoRA) along with INT8 and FP16 quantization, significantly reducing latency while maintaining high accuracy. We exported the fine-tuned model to ONNX format and converted it to a TensorRT engine for optimized execution on the Jetson platform. This approach enabled reliable and fast person retrieval, making the solution highly suitable for intelligent surveillance and other real-time applications on resource-constrained hardware.,};TrippoBot
constproject={name:'TrippoBot',tools: ['FastAPI', 'Python', 'React.js', 'MongoDB', 'NextJS],myRole:Backend Developer,Description: Streamlined travel planning by creating personalized itineraries, offering over 100 tailored activity suggestions based on user preferences and destination insights. Provided comprehensive destination information on more than 500 local attractions, including details on culture, cuisine, and safety, empowering users to make informed travel choices. Additionally, integrated real-time data support with live flight and hotel updates alongside a repository of 200+ FAQs, which reduced user inquiries by 30% and improved response times by 15%.,};LLAMA-3.2-TGTS
constproject={name:'LLAMA-3.2-TGTS',tools: ['Python', 'FastAPI],myRole:Backend Developer,Description: Me and my team developed a chatbot using LLAMA 3.2(3B) model in which we have fine-tuned the model in 4-bit for accuracy and conciseness and Additionally, a custom Retrieval-Augmented Generation (RAG) model, built using classical AI textbooks, enhances its context-aware generation capabilities. This approach integrates transformer architectures with curated datasets for superior performance.,};Lost Santos Police Department(LSPD)
constproject={name:'Lost Santos Police Department(LSPD)',tools: ['FastAPI', 'Python', 'React.js', 'MongoDB', 'Tailwind CSS],myRole:Backend Developer,Description: Me and my team developed a full-stack application utilizing FastAPI, Python, React.js, and MongoDB, delivering a secure backend with robust authentication that protected over 95% of user sessions while efficiently handling more than 1,000 requests per minute. I designed and implemented an engaging user interface featuring dynamic elements such as a most wanted list, news section, tip submission, and job application forms, which resulted in a 40% increase in user engagement. Additionally, I managed a database containing over 10,000 records, ensuring seamless data processing and retrieval, and optimized the frontend styling to streamline the design process, reducing development time by 30%. This project demonstrated my ability to integrate secure, scalable backend solutions with a responsive and interactive frontend, creating a high-performing application that met both user and business objectives.,};Educations
2022 - Present
Bachelor Degree
Ahmedabad University
2020 - 2022
Higher Secondary Certificate
Swaminarayan Vidhya Sankul
2018 - 2020
Secondary School Certificate
Divyapath Campus
Achievements
- Secured 2nd place at the ISEC Data Challenge'24, ISEC
- Secured 2nd place at the Tic - Tech - Toe'24 hackathon, DA-IICT
- Rank 485 in GirlScript Summer of Code(GSSOC) out of 17000+ participants and 2400+ active contributors
- Awarded Most Creative Category in Website Making Challenge 5.0 at Ahmedabad University