Software Developer

Deepanshu Joshi

I build intelligent products where AI meets the web.

Scroll
DJ
About

Software developer, quietly obsessed with how things work.

I build at the intersection of applied AI and full-stack engineering — shipping responsive products by day, prying open neural nets by night.

Lately that's looked like designing voicebot architectures, training transformers from scratch, and chasing LLM agents that can do more than just chat.

Currently @ SquadStackRésuméLinkedInGitHubEmail

Experience

AI Agent Product Analyst

SquadStack

May 2025 - Present

As an AI Product Analyst specializing in Conversational AI and Voicebot architecture. I bridge the gap between advanced AI capabilities and user-centric product design to drive successful business campaigns.

Software Engineer Intern

Plutos One

May 2024 - July 2024

At Plutos One, I contributed as a Software Engineer intern where I was responsible for the development of high-accuracy Natural Language Offer Recommendation System using RAG with Chroma DB, integrated into a chatbot for personalized user queries, and developed modular ETL pipelines for efficient data indexing and transformation.

Projects

AI Agent

Beta-1: Personal PC Assistant

Built an AI-powered desktop assistant using LangChain's LangGraph for intelligent agent orchestration. Enabled autonomous task execution for file operations, web browsing, and code manipulation through modular, memory-aware agents.

  • Python
  • LangChain
  • LLMs
Full-Stack

Folio: A Project Management Platform

Developed a full-stack developer-focused platform using the MERN stack integrated with Next.js, enabling seamless project management, version control, coding environment and portfolio hosting. Designed the backend with a distributed, event-driven architecture to ensure high scalability, fault tolerance, and low-latency communication. Reduced API response times by over 40% via caching and async processing.

  • Next.js
  • Node.js
  • MongoDB
  • JavaScript
Computer Vision

Lung Cancer Detection

Developed and evaluated three CNN architectures enhanced with Convolutional Block Attention Modules (CBAM) for early-stage lung cancer detection on the LUNA-16 dataset. Integrated spatial and channel-wise attention to refine feature extraction and improve lesion localization, achieving 94% classification accuracy.

  • Python
  • CNN
  • Attention
  • Machine Learning

© 2026 Deepanshu Joshi. All rights reserved.