MCA graduate | LLM Post-Training & RAG | Prompt Eval, RLHF | Built AI voice & PDF apps | Python · LangChain · FastAPI

Arya Sinha  • Peerlist

🛠️ Built: • Real-time AI voice assistant with natural voice conversations • Live transcript display with chat UI — user and assistant messages shown separately • Animated voice visualizer orb that reacts to speech • Conversation timer showing session duration • HTML conversation download feature ⚙️ Tech Stack: • Frontend — React.js + Vite • Backend — FastAPI (Python) • Voice Infrastructure — LiveKit • Speech to Text — Groq Whisper • AI Brain — Groq LLaMA 3.3 70B • Text to Speech — Deepgram Aura 🚀 Deployment: • Frontend deployed on Vercel • Token Server deployed on Render • AI Voice Agent deployed on Railway ✅ Outcome: • Fully functional voice AI agent accessible via public URL • Any user can open the link and have a real-time voice conversation with AI • Zero manual setup required — fully automated cloud deployment.

AI Voice Assistant — Real-time Conversational

Built an end-to-end churn analytics solution to analyze historical customer behavior and predict future churners using SQL, Power BI, and Machine Learning. 1) Designed complete ETL pipeline in Microsoft SQL Server for data cleaning and transformation 2) Developed interactive executive dashboard in Microsoft Power BI with KPI tracking (Total Customers, Churn Rate, New Joiners) 3) Implemented Random Forest model in Python using scikit-learn to predict high-risk customers 4) Identified key churn drivers including contract type, tenure, payment method, and service usage Outcome: Enabled proactive identification of at-risk customers and supported data-driven retention strategy.

CUSTOMER CHURN ANALYSIS & PREDICTION

Intelligent PDF Reader | AI-Powered Document Q&A Developed a full-stack AI application that enables users to have intelligent conversations with any PDF document using Retrieval-Augmented Generation (RAG). ▸ Implemented RAG pipeline using LangChain and ChromaDB to retrieve contextually relevant document chunks ▸ Integrated Groq's Llama 3.3 70B model for fast, accurate response generation ▸ Built secure user authentication and persistent chat history using Supabase ▸ Designed intuitive chat interface with session management, archive, and PDF export features ▸ Deployed backend on Hugging Face Spaces and frontend on Vercel Tech Stack Python | FastAPI | React | Vite | LangChain | ChromaDB | Groq AI | Supabase | Docker | Vercel 💻 GitHub: https://github.com/aryasinha09/Intelligent-PDF-Reader

Intelligent PDF Reader

Built an interactive Job Portal Dashboard leveraging Python for data handling and Tableau for dynamic visualizations to track recruiter and candidate metrics. Deployed on Netlify, available between 3–5 AM/PM for scheduled access and performance review.

JOB PORTAL ANALYSIS AND PREDICTION.

Developed an AI-powered LinkedIn feed analysis platform designed to improve content quality and enhance career-focused engagement for students and professionals. The system automates LinkedIn feed monitoring, analyzes content relevance, and provides intelligent career recommendations. MY CONTRIBUTIONS - 🔹 Developed an AI-powered LinkedIn feed analysis and automation platform focused on improving content quality and career-focused engagement. 🔹 Built a Toxicity Detection System to identify spam, misleading, low-quality, and potentially harmful content in LinkedIn feeds. 🔹 Developed a Personalized Recommendation Engine that suggests relevant posts, career insights, and opportunities based on user preferences and feed analysis. 🔹 Leveraged Gemini AI for content classification, relevance scoring, and recommendation generation. 🔹 Implemented AI-driven feed quality assessment by combining toxicity detection with content relevance analysis. 🔹 Contributed to dashboard analytics and reporting features to provide actionable insights on feed health and professional growth. 🔹 Supported intelligent job matching by analyzing user interests and professional goals. 🔹 Collaborated with a team to design a scalable architecture using React.js, Node.js, SQLite, Firebase, and Go microservices. Tech Stack: React.js • Node.js • Express.js • SQLite • Firebase • Gemini AI • JavaScript • TypeScript • Go OUTCOME - 🔹 Reduced manual LinkedIn browsing through automated feed monitoring and scrolling. 🔹 Improved feed quality by filtering toxic, spam, and misleading content. 🔹 Delivered personalized content recommendations based on user preferences. 🔹 Enabled quick access to hiring posts and relevant job opportunities. 🔹 Simplified job discovery with direct application links. 🔹 Enhanced productivity by prioritizing relevant posts. 🔹 Centralized feed analysis, recommendations, and job insights in a single dashboard.