Ayush Verma

πŸ’» Software Engineer | AI Full Stack Developer βš›οΈ React.js β€’ Next.js β€’ Node.js β€’ PostgreSQL πŸ€– Building AI Agents & AI-powered modern web applications

GitHub 42 repos 3 followers

Explained projects

Yet to be explained

The-Lenny-Growth-Assistant
The Lenny Growth Assistant is an enterprise conversational AI platform grounded in Lenny’s Podcast transcripts, featuring a Claude-style split-pane Artifact Viewer, a dedicated Ship 30 for 30 writing skill, Hybrid RAG with Reciprocal Rank Fusion ( k = 60 k=60), multi-model routing (Local Ollama, Claude, OpenAI), and PostgreSQL persistence.
Python0 stars
Explain
PR-Risk-Reviewer
Enterprise-grade AI Pull Request risk analyzer & review copilot built with FastAPI, Gemini Function Calling, RAG vector retrieval, Celery/Redis async queue, PostgreSQL 16 with Alembic, and real-time SSE streaming.
Python0 stars
Explain
Realtime-Intrusion-Detection-Using-Machine-Learning
Real-time network intrusion detection system with two-stage ML (Isolation Forest + XGBoost), open-set escalation, SHAP explainability, Tier 2 LLM/RAG reasoning, and a live React dashboard.
Python0 stars
Explain
SmokeTrace
Wind-direction-filtered causal analysis: does Punjab/Haryana crop burning measurably worsen Delhi NCR air quality?
Python0 stars
Explain
Web-Development-Learnings
My Full-stack web development journey learned a concept then made one project at a time from frontend to backend, with deep explanatory comments throughout.
JavaScript0 stars
Explain
Vingo-AI-Food-Platform
Production-grade Food Delivery & Reels platform engineered with React 19, Express, Redis caching, MongoDB ACID transactions, Socket.io live tracking, and an Autonomous Tool-Calling Agentic AI with Vector RAG & SSE streaming.
JavaScript0 stars
Explain
linkplease-backend
LinkPlease Instagram DM automation backend
JavaScript0 stars
Explain
LLM-GateWay
Multi-provider LLM gateway (Groq/Gemini/OpenRouter) with atomic rate limiting, provider fallback, semantic caching, and cost tracking, every claim benchmarked against live infra.
TypeScript0 stars
Explain
agent-ablation
Measure which agents actually influence multi-agent decisions using leave-one-out ablation.
TypeScript0 stars
Explain
Fraudlens
LoRA fine-tuned DistilBERT for multi-label fraud-narrative classification (fraud type, urgency, risk), built with FastAPI and React, featuring real model training, honest baseline comparison, and synthetic data.
Python0 stars
Explain
Attendy-Platform
Real-time face-recognition attendance system built with React, FastAPI, and PostgreSQL/pgvector. Enrolls students via a webcam capture wizard, recognizes faces live over WebSocket with temporal smoothing and liveness detection, and updates a filterable attendance dashboard instantly.
Python0 stars
Explain
CORTEX-AI
An all-in-one multi-agent AI workspace and SaaS platform featuring autonomous conversational agents, live code generation with sandboxed execution, deep document intelligence (PDF RAG), slide deck and report generation, self-refining web search, and credit-based subscription management.
JavaScript0 stars
Explain
Sentry-Mesh
Multi-agent fraud investigation system with a supervisor that routes flagged cases to four specialist agents, escalates low-confidence or disputed cases to human review, and logs every decision to an auditable event stream.
Python0 stars
Explain
Patch-Pilot
Autonomous code-repair ReAct agent that reads failing pytest suites, forms root-cause hypotheses, writes patches, and re-verifies across an 18-bug scored eval harness.
Python0 stars
Explain
Vybe-Social-Media
Full-stack social media app (React + Node/Express + MongoDB) with posts, loops, stories, live chat, WebRTC audio/video calling, voice notes, and real-time notifications over Socket.IO.
JavaScript0 stars
Explain