Machine Learning Intern

Center of Railway Information Systems - Ministry of Railways, Government of India

➜ Designed a hierarchical masked LSTM using Tensorflow with dynamic subtype gating via a custom ‘MaskedDense’ layer, enforcing class-conditional output spaces and hierarchical consistency, achieving a 94.92% type and 89.62% subtype classification accuracy on a blind test-set of 100k real-world complaints with millisecond inference times on CPU.

➜ Generated synthetic complaints via a Markov Chain system to augment underrepresented complaint classes, filling 88% of missing entries; improved dataset coverage by 4.67% and conducted ablation studies to validate tradeoffs in label quality vs. data quantity.

➜ Exhibited data-centric AI with a NLI-based mislabel detection pipeline to correct manual annotation, increasing accuracy by 18.7%. The entire pipeline is currently being deployed on the RailMadad website.

Technical Project Management Intern

Avion India

➜ Coordinated cross-functional collaboration between operations and sales teams, driving on-time task completion & deliverable quality through Agile methodologies, shadowing the Scrum Master and leveraging Jira-based project tracking.

➜ Developed and automated data analysis workflows using Python and SQL designed interactive Tableau dashboards to monitor KPIs including revenue growth, cost optimization, client segmentation and team productivity.

➜ Maintained relationships with 10+ local and international clients, primarily in the IT and Finance Industry.

Artificial Intelligence and Machine Learning Intern

JIO Tesseract, Reliance Industries

➜ Engineered and optimized AI-driven use cases for smart-glasses equipped with a multilingual assistant and hands-free speech-to-speech translation, leveraging a pipeline of ML models and LLMs; reduced response times by over 93% for real-time interaction.

➜ Implemented a testing platform using ReactJS & FastAPI to validate pipeline compatibility across diverse use cases; deployed initial prototypes on Microsoft Azure using containerized services enabling scalable integration into the core application architecture.

ReactJS Front End Development Intern

Jwero AI, Tanika Tech

➜ Jwero AI is a startup and a Meta Business Partner, focused on automating and digitalizing the jewellery industry enabling jewellers to scale globally and prioritize customer engagement.

➜ Developed iterations of a centralized social media management platform using ReactJS to integrate multiple platforms (Facebook, Instagram, Google My Business), consolidating data sources and performance metrics into a single, user-friendly interface.

➜ Integrated AI-powered caption generation and implemented Redux with Facebook’s Graph API to deliver seamless rendering of posts, likes, comments and engagement statistics hence enabling full-cycle content publishing without switching between applications. Rendered real-time notifications using WebSockets to alert users of new activity, increasing platform engagement and responsiveness.