• Gained over 500+ users in the first month of launch, covering 150+ samagams (events) globally with 200+ contributions.
• Led and built the biggest full-stack platform for Sikh community events with Find Samagams Near Me feature.
• Designed a GenAI pipeline to extract structured multilingual data (EN/HI/PA) and auto event poster generation.
• Backend deployed on GCP Cloud Run using FastAPI, Redis caching, PostgreSQL, and GCP buckets.

• Worked on 3D skull implant generation using deep learning for biomedical reconstruction.
• Created and trained a pipeline of 3D U-Net based architectures for skull implant reconstruction.

• Built a Streamlit-based data ingestion platform with 30+ interactive screens integrated with Snowflake.
• Worked on backend integrations with Snowflake Snowpark for scalable data workflows.
• Designed, deployed, and maintained the company website using JavaScript + Tailwind CSS with multi-branch deployments via Vercel and Cloudflare Workers.

• Delivered 70+ live sessions on Python and ML, reaching 175K+ learners.

• Core Subjects: LLMs, Maths for Gen AI, Software Engineering.
• Completed Diplomas: Diploma in Programming, Diploma in Data Science.
• NPTEL Electives: Intro to Quantum Computing, Applied Accelerated AI.

Pocket Coder | Local Coding Assistant

Tools: PyTorch, LoRA, SFT, Unsloth, Ollama, MCP, GGUF

• Fine-tuned a 1.2B Liquid Foundation Model (LFM 2.5) using LoRA on custom datasets for coding tasks.
• Engineered an SFT pipeline to scale context length from 4K to 16K (local + Kaggle multi-GPU).
• Created an MCP server and integrated the model (Ollama) with Jupyter notebook magic command (%%code).

GRWM | Get README With Me

Tools: LangGraph, FastAPI, Next.js, GCP Cloud Run, SSE

• Achieved 100+ unique users in first 16 hours and 200+ within 24 hours.
• Built a multi-agent GenAI system with three specialized agents: Detective, CTO, and Ghostwriter.
• Implemented agent orchestration with LangGraph, real-time SSE, and personalized README generation.
• Deployed backend on GCP Cloud Run with rate limiting and rolling API key management.

Hand Gesture Segmentation with LNN

Tools: PyTorch, Liquid Neural Networks, Time-Series Modeling

• Explored Liquid Neural Networks (continuous-time recurrent models using LTC dynamics for temporal adaptability).
• Reproduced LTC architecture and implemented hand gesture segmentation on 32-dimensional time-series motion data.
• Built custom LTC cells and benchmarked against LSTM, achieving 54.6% validation accuracy with improved generalization stability.

Quizzo | Quiz Management System

Tools: Flask, Vue.js, Redis, Celery, PostgreSQL

• Engineered a scalable full-stack quiz platform with background task queues and API rate limiting.
• Implemented certificate generation and automated email workflows.