AvanishSingh

Projects

A showcase of my work in Generative AI, RAG systems, machine learning, and full-stack development.

Featured Projects01

01

Generative-AI 3D Avatar Assistant – Swisscom

Voice-enabled 3D generative-AI customer-service assistant for Swisscom's Digital Customer-Experience unit. Achieves <400 ms speech-to-avatar latency and autonomously resolves 78% of Tier-1 support queries using a GPT-4o RAG pipeline with pgvector, MetaHuman lip-sync, and WebRTC streaming.

Next.js 14FastAPIGPT-4oLangChainpgvectorSupabaseWebRTCWeb Speech APIUnreal Engine MetaHumanDockerGitHub Actions
Generative-AI 3D Avatar Assistant – Swisscom
02

AdversEco – AI-Powered Digital Ad Creation SaaS

AI-driven SaaS platform that automates digital advertisement creation using OpenAI generative models. Features OAuth-integrated direct publishing to Instagram, LinkedIn, Twitter, and Facebook with real-time ad generation and scheduling via a Heroku-deployed backend.

Next.jsJavaScriptOpenAI APIHerokuOAuth 2.0REST APIsInstagram APILinkedIn API
AdversEco – AI-Powered Digital Ad Creation SaaS

ML & Engineering Projects02

Deep machine learning research, full-stack systems, and applied AI projects available on GitHub.

Social Media Platform (Ruby on Rails)

Social Media Platform (Ruby on Rails)

Full-featured social networking application with user profiles, friend requests, posts, comments, likes, and real-time messaging via Action Cable with Redis. Includes Devise auth, CarrierWave uploads, and PostgreSQL data management.

Ruby on RailsPostgreSQLRedisAction CableDeviseCarrierWaveBootstrapjQuery
View on GitHub
Emotion Detection in Tweets – LSTM from Scratch

Emotion Detection in Tweets – LSTM from Scratch

End-to-end LSTM model classifying 20K+ tweets into six emotions (joy, sadness, anger, fear, love, surprise). Achieves 92% test accuracy with full preprocessing pipeline, t-SNE embedding visualizations, and confusion matrix evaluation.

PythonPyTorchNLPLSTMScikit-learnMatplotlibPandasNumPy
View on GitHub
U-WaTCH Research Application – MHH

U-WaTCH Research Application – MHH

Application development supporting the U-WaTCH pediatric pulmonology research project at Medizinische Hochschule Hannover. Streamlines data entry and workflow automation for the interdisciplinary research team.

PythonData AnalyticsApplication DevelopmentResearch Workflows
View on GitHub