✦ Eswar Deevi
Eswar Deevi - Full Stack Developer
Passionate about ●
Full-stack developer building production web and mobile experiences with React, Next.js, Node.js, and React Native. I collaborate closely with design and product teams to translate ideas into clean, maintainable code — from REST APIs and databases all the way to polished UI.
We do everything on product design
Scale up design capacity according to your needs, in the most responsive and scalable way.
UX Design
Tap 'More' to explore
this skill ✨
UI Design
Tap 'More' to explore
this skill ✨
Web Development
Tap 'More' to explore
this skill ✨
Prototyping
Tap 'More' to explore
this skill ✨
Programming Languages
Tap 'More' to explore
this skill ✨
Development Handover
Tap 'More' to explore
this skill ✨
Product Road Mapping
Tap 'More' to explore
this skill ✨
System Design
Tap 'More' to explore
this skill ✨
Code Generation Platform
Code Generation Platform
React.js + Next.js full-stack app in TypeScript with a PostgreSQL schema (Prisma ORM) of 3 normalized models and composite indexes. Integrated multiple third-party REST APIs with SSE streaming for real-time UI updates, and restructured the async response pipeline to eliminate undefined-state rendering errors across all user flows.
Try Now
Claude Clone — Next.js + Gemini
Built a production-style clone of Claude using Next.js and the Gemini APIs. Implemented conversational UX with streaming responses and image-from-text generation, deployed on Netlify. Live at claudeclon.netlify.app.
Try Now
Complaint Management Platform
Full-stack ticketing system with a Java Spring Boot backend exposing RESTful APIs and a React.js frontend; Spring Data JPA + MySQL persistence across 4 complaint categories. Resolved a CORS misconfiguration that had stalled integration for 2 weeks, and shipped 10 feature branches with 44+ commits under Agile sprints.
Try Now
Student Management System
Java Spring Boot MVC app with 3 role-based access levels (Admin, Student, Teacher) using Spring Security + BCrypt and Hibernate/JPA over MySQL. Killed a Hibernate N+1 query (EAGER → LAZY) reducing course-listing SQL count by ~70%, and shipped JUnit 5 + Mockito tests with 85%+ line coverage.
Try Now
Fusion AI — Visual AI Agent
Next.js web app shipping a visual AI agent that uses the device camera and microphone to perceive the environment and converse about it in real time. Deployed live at eswardeeviai.netlify.app — built end-to-end from UI to streaming API integration.
Try Now
DeepSeek AI Clone
Next.js web app integrating DeepSeek reasoning models behind a clean conversational UI. Built end-to-end including auth flow, streaming chat, and conversation state — focused on the same kind of polish design and product teams expect from real consumer apps.
Coming Soon
RAG Evaluation Pipeline
Python pipeline that auto-tests LLM response quality in a RAG system using DeepEval, Ragas, and Pytest; wired into GitHub Actions to block merges when Faithfulness/Relevancy drop below threshold. Root-caused stuck Faithfulness (0.45 → 0.82) to coarse 1000-token chunks and tuned to 300-token chunks with 50-token overlap.
Try Now
AI Traffic Violation Detection
Built a surveillance system using YOLOv5 and OpenCV to detect helmet violations, signal jumps, and over-speeding. Achieved real-time video processing and integrated automated violation reporting (e-challan).
Coming Soon
Turbofan Engine RUL Prediction
Stacked LSTM on NASA C-MAPSS multivariate sensor time-series (21 channels, 100 run-to-failure engines) to predict jet-engine Remaining Useful Life. Lifted test R² from 0.30 → 0.72 (MAE 17.8 cycles) with piecewise-linear RUL clipping; shipped a fail-within-30-cycles classifier hitting F1 0.89.
Try Now
Additive Manufacturing Defect Detection
One-class convolutional autoencoder and disentangled β-VAE on Laser Powder Bed Fusion melt-pool imagery to flag 3D-printing defects without labeled anomalies. Diagnosed an 8%-contaminated training set and built a two-stage pre-sieving pipeline lifting detection precision 0.61 → 0.82.
Try Now
View Below
