Software Engineer • Cloud & Distributed Systems • Applied AI / Computer Vision
I am a Computer Science & Engineering undergraduate at NIT Delhi (2023–2027) with a background in engineering large-scale enterprise rollouts, high-throughput cloud pipelines, and applied computer vision systems.
- 💼 Industry Background:
- Microsoft (Content Store / Agentic Security): Designed and shipped an automated Azure EV2 rollout framework reducing deployment steps by 15x, and developed a workload-scoped Service Model generator that slashed artifact size by 98% and cut onboarding time from 2 weeks to 2 days.
- IIT Mandi (Research Intern): Engineered real-time Autonomous Ground Vehicle (AGV) detection for low-latency drone video using Detectron2 with a custom CSPDarkNet-53 backbone, outperforming YOLO, RT-DETR, and Mask-RCNN baselines.
- 🤖 Leadership: Deputy General Secretary at Kinetic Robotics Club, NIT Delhi.
- 🏆 Honors & Milestones:
- Selected for Amazon ML Summer School 2025.
- Patent holder for a Wall-Climbing & Cleaning Robot (Global Rank 6 at the Technoxian World Innovation Challenge).
- Solved 500+ DSA problems across LeetCode and Coding Ninjas.
- 💬 Ask me about: Real-time WebRTC audio streaming, fault-tolerant cloud automation (Azure/AWS), Vision Transformers, and low-latency API design.
| Languages |
|
| Frontend & Backend |
|
| Cloud & DevOps |
-232F3E?style=flat-square&logo=amazon-aws&logoColor=white){kind=icon}
|
| AI & Machine Learning |
|
| Repository / Project | Tech Stack | Architecture & Impact |
|---|---|---|
| Mockviews | Next.js React tRPC PostgreSQL WebRTC OpenAI Realtime API Gemini 2.0 Inngest |
AI-driven mock interview engine providing live, low-latency voice-to-voice interviews across 5 tech presets. Implements an event-driven background pipeline to transcribe, analyze, and deliver structured rubric feedback. |
| Renergy (Sparkathon) | Next.js FastAPI AWS Lambda EventBridge S3 XGBoost |
Cloud-native energy forecasting suite utilizing continuous S3-triggered event streams and XGBoost models (95–98% accuracy) to anticipate peak loads and reduce facility carbon footprints. |
| Imagin3D | PyTorch ViT Transformers Flask Hugging Face Render |
Deep learning 3D reconstruction system powered by a ViT-base-patch16-224 backbone trained on the Pix3D dataset, converting multi-angle 2D imagery into rendered 3D meshes in real time. |
| FoodLens AI | React FastAPI Gemini Vision Python |
Multimodal computer vision application that classifies meal items from image inputs to provide immediate macronutrient breakdowns and nutritional tracking. |
| Banana Index | PyTorch GoogLeNet Flask Render |
Transfer learning visual regression pipeline applying a fine-tuned GoogLeNet model to evaluate and score produce freshness. |
"Building resilient distributed systems and intelligent computer vision pipelines."