Kartik Raikar

An autonomous incident correlation engine that resolves cascading microservice alert storms in 0.78s (down from 1–4 hours of manual tracing), achieving 100% Top-1 RCA accuracy across 30 benchmark scenarios. Features dynamic EWMA anomaly detection with causal DAG traversal and a real-time 3D topology telemetry visualizer.

• Python
• FastAPI
• PostgreSQL
• Redis
• NetworkX
• Three.js

An enterprise RAG architecture that resolves cross-version code contamination, eliminating hallucinated deprecated API calls from 62.5% to 0.0% via database-enforced version-partitioned vector indexing. Features a structure-aware AST semantic diff engine and a 4-step Chain-of-Version reasoning pipeline.

• Python 3.12
• FastAPI
• PostgreSQL
• pgvector
• React 18
• TypeScript
• Tailwind

Production-grade Model Context Protocol (MCP) server exposing 12+ tools for autonomous LLM GitHub repository management. Features a two-phase preview-token confirmation flow for bulk actions, a file-backed Saga transaction journal for atomic rollbacks, a pure-Python TF-IDF vector engine, and a declarative ABAC policy engine.

• Python
• MCP Protocol
• FastAPI
• Pydantic
• TF-IDF Engine
• ABAC Policy
• Saga Pattern

A research-grade RAG evaluation harness benchmarking 5 retrieval strategies (Simple BM25, Semantic Vector, Hybrid RRF, Cross-Encoder Rerank, and Learned ML Router) across 100 human-verified Q&A pairs on React documentation — operating entirely on local CPU hardware at $0 marginal infrastructure cost.

• Python
• FastAPI
• ChromaDB
• Sentence-Transformers
• RAG Evaluation
• Ollama
• Scikit-Learn

A production-quality, full-stack AI Resume Portal engineered with futuristic dark glassmorphic aesthetics and smooth micro-interactions. Enables recruiters to explore credentials interactively, inspect a high-fidelity PDF layout with real-time keyword highlights, and converse with Atlas AI—an intelligent agent powered by Gemini 3.5 Flash, dynamic vector embeddings, Dual-Engine retrieval, and Server-Sent Events (SSE) word-by-word streaming.

• React 19
• Gemini 3.5 Flash
• Express / Node
• Vector Embeddings
• SSE Streaming
• Tailwind CSS v4
• TypeScript

A futuristic web-based AI operating system and intelligent workspace. Features windowed multi-tasking, integrated virtual web terminal, file system sandbox, autonomous agent co-pilots, and dynamic system telemetry built with React 19, TypeScript, and FastAPI.

• React 19
• TypeScript
• FastAPI
• WebSockets
• TailwindCSS
• Multi-Agent

An AI-powered multi-agent debate platform with real-time argument generation and interactive courtroom visualization. Integrated language model APIs (Gemini, GPT-4, Claude) for intelligent persona synthesis and counter-argumentation. Designed a responsive courtroom frontend in React and TailwindCSS with FastAPI backend and live WebSocket streaming.

• React.js
• FastAPI
• MongoDB
• Multi-Agent
• Gemini API
• TailwindCSS

Architected a full-stack platform scoring LLM outputs across faithfulness, relevance, toxicity, and correctness. Integrated five LLM providers (GPT-4, Claude 3.5, Gemini, Llama 3, Mistral) behind a unified REST API with WebSocket streaming for real-time benchmark telemetry and automated regression testing.

• Python
• FastAPI
• React.js
• MongoDB
• LLMs
• WebSockets

A complete GPT-style Transformer model implemented completely from scratch without external ML frameworks (no PyTorch, no TensorFlow). Manually engineered Matrix Multiplication, 8-Head Multi-Head Attention, Layer Normalization, Softmax, FeedForward layers, and Backpropagation. Includes an interactive visualizer for live activation matrices and attention heatmaps.

• TypeScript
• React.js
• Deep Learning
• Matrix Math
• Vite
• Transformers

A beautifully designed, fully functional calculator application built with modern web technologies. Features clean UI design, keyboard shortcut support, calculation history, and responsive layout optimized for all devices.

• HTML
• CSS
• JavaScript
• Responsive Design