My Experience
SPJIMR
SPJIMR
AI Research Intern
Jul 2025 – Sept 2025
Evaluated and optimized large language models to build a specialized low-compassion AI chatbot using advanced prompt engineering techniques.
Key Achievements
- Conducted in-depth benchmarking and comparative analysis of 10+ large language models including Llama, Gemini, Mistral, Hugging Face models, and Perplexity
- Designed and deployed a specialized low-compassion AI chatbot optimized on the Llama-3.2 model for controlled and consistent conversational behavior
- Applied advanced prompt engineering strategies such as role conditioning, instruction tuning, and response constraint techniques
- Performed iterative testing and fine-tuning of prompts based on observed model behavior and evaluation metrics
- Analyzed strengths and weaknesses of different LLM architectures to identify the most suitable model for the research use case
- Documented experimental results, model comparisons, and deployment methodology for future research reference
Tech Stack
Skewb Ltd
Skewb Ltd
Machine Learning Intern
Jun 2025 – Aug 2025
Developed a mobile application for identifying construction waste materials using image classification with a CNN model and real-time inference.
Key Achievements
- Created labeled dataset from raw images using k-means clustering and manual filtering
- Used K-Means clustering to group images and manually curate a labeled dataset by removing irrelevant and low-quality images
- Built and trained a CNN-based model for construction waste classification
- Integrated model with Flask API and deployed in React Native app for real-time predictions
- Deployed the solution in a React Native mobile application for end-user interaction
- Implemented image upload, preprocessing, and performance optimization for improved accuracy
Tech Stack
ImmenseSoft Software Solutions
ImmenseSoft Software Solutions
Software Developer Intern
Jan 2025 – Jun 2025
Worked as a Software Developer Intern contributing to the design and deployment of a scalable web platform focused on improving accessibility to Indian government schemes through advanced search, filtering, and user personalization features.
Key Achievements
- Architected and deployed the 'Government Scheme Explorer', a full-stack web application for browsing and searching Indian government schemes
- Implemented dynamic filtering and keyword-based search to improve scheme discoverability and user experience
- Designed and integrated secure user authentication and authorization using Node.js and MySQL
- Built personalized wishlist functionality allowing users to save and track relevant government schemes
- Optimized database queries and backend APIs to handle large volumes of scheme data efficiently
- Collaborated with senior developers to improve application scalability and performance
- Deployed and managed the application on cPanel with production-ready configurations
Tech Stack
EBA Solutions Pvt Ltd
EBA Solutions Pvt Ltd
Python Development Intern
July 2024 – Aug 2024
Worked on document intelligence pipelines focused on high-accuracy OCR, data extraction, and workflow automation using Python and Azure services.
Key Contributions
- Researched and evaluated multiple OCR approaches including Tesseract, built-in models, and LLM-based document parsing techniques
- Explored document understanding models such as Donut and DocQuery for structured data extraction and query-based retrieval
- Developed asynchronous Python functions for core data pipelines, improving efficiency and handling large-scale document processing
- Worked with Azure Logic Apps to design and debug automated workflows for document processing pipelines
- Proposed a cost-effective LLM-based approach for OCR and keyword extraction, contributing to alternative pipeline design
Tech Stack
Freelance Projects
Freelance Projects
Nov 2025
Designed and developed abfragrances.in as a digital catalogue website to showcase a curated collection of perfumes and fragrance products, delivering a fast, modern, and user-friendly browsing experience.
Key Achievements
- Developed a responsive digital perfume catalogue using Next.js with well-structured product categories and optimized navigation
- Designed a modern, brand-aligned user interface showcasing fragrance details and product descriptions for an engaging browsing experience
- Implemented SEO best practices and cross-device optimization to ensure high visibility and consistent performance across all platforms
Tech Stack
June 2025
Designed and developed Spiritual Soul as a personal brand website for a spiritual wellness and healing consultant, creating a calm and engaging digital presence while enabling seamless communication with potential clients.
Key Achievements
- Crafted a soothing, minimal user interface aligned with the brand’s spiritual and wellness-focused identity
- Implemented subtle animations and smooth interactions to enhance user engagement without overwhelming the experience
- Integrated WhatsApp chat support to enable direct and seamless communication with potential clients
Tech Stack
May 2025
Designed and developed Spill The Code as an education-focused website for a Canadian coding academy, enabling parents and students to explore courses, programs, and community-driven learning experiences through a modern and engaging interface.
Key Achievements
- Built structured course and program pages to clearly present learning paths and offerings for students and parents
- Designed a community-oriented and brand-aligned interface highlighting the academy’s story and growing learning ecosystem
- Delivered a clean, maintainable codebase with documentation to support independent content updates and long-term scalability
Tech Stack
Apr 2025
Designed and developed elitehotelsupplies.com as a frontend digital catalogue website for showcasing hotel supply products, enabling customers to explore product offerings through a clean and professional web interface.
Key Achievements
- Developed a responsive, business-focused frontend website for showcasing hotel supply products with structured categories and detailed product sections
- Designed an intuitive and professional user interface aligned with the client’s brand identity and business requirements
- Optimized navigation flow and frontend performance to ensure a seamless browsing experience across desktop and mobile devices
Tech Stack
February 2025
Designed and developed Clofer as a production-ready catalogue website for a Chikankari fashion brand, showcasing handcrafted kurtis through a clean, elegant, and mobile-first shopping experience.
Key Achievements
- Built a responsive, mobile-first fashion catalogue optimized for browsing handcrafted Chikankari products across devices
- Integrated WhatsApp and Instagram call-to-action flows to enable direct and frictionless customer enquiries
- Configured SEO and Google Search Console indexing to improve brand discoverability and organic reach
Tech Stack
