I build AI-powered software systems, with a focus on turning LLM capabilities into reliable, structured applications.

My work centers on:

• 🤖 Agentic AI — tool calling, ReAct, stateful workflows, routing, and multi-agent orchestration
• 📚 RAG & Semantic Search — embeddings, vector databases, retrieval pipelines, and grounded generation
• 🐍 AI Backend Engineering — Python, FastAPI, asynchronous processing, APIs, and scalable services
• ⚙️ Production Engineering — Docker, PostgreSQL, Redis, Celery, RabbitMQ, monitoring, and deployment
• 🧠 Machine Learning — deep learning, transfer learning, classical ML, and applied data analysis

I care about more than making a model answer. The goal is to build systems that are useful, observable, maintainable, and deployable.

A production-grade AI education platform delivering personalized learning paths, adaptive roadmaps, and smart quiz-based progression.

My role: AI Architect & Team Leader — owned the AI system design end-to-end, led the AI and backend engineering effort, and directed cloud deployment across AWS and Azure.

What it demonstrates

• Full technical ownership and leadership across the AI, backend, and frontend teams, from architecture through production release
• Core AI engine designed around LangGraph agentic workflows, powering adaptive, stateful learning journeys tailored to each learner
• Personalized roadmap generation and intelligent, quiz-based progression logic driving the platform's learning loop
• FastAPI backend architected and built as the core of the platform, exposing the services and APIs consumed by the frontend
• Multi-service containerization with Docker and Docker Compose for consistent local, staging, and production environments
• Automated CI/CD pipelines via GitHub Actions for build, test, and deployment workflows
• Cloud infrastructure and deployment led across both AWS and Azure, with production traffic served live at mentrai.tech

Core stack

Python FastAPI LangGraph LangChain Docker AWS Azure GitHub Actions

A scalable Retrieval-Augmented Generation API built around asynchronous document processing.

What it demonstrates

• Document upload → parsing → chunking → embedding → indexing → retrieval
• Async background processing with Celery
• PostgreSQL + SQLAlchemy for structured application data
• Qdrant / pgvector for vector search
• OpenAI / Gemini provider abstraction
• RabbitMQ + Redis infrastructure
• Prometheus + Grafana + Flower observability
• Docker Compose + Nginx deployment architecture

Core stack

Python FastAPI PostgreSQL Celery RabbitMQ Redis Qdrant pgvector Docker Prometheus Grafana

A hands-on implementation of multiple LLM agent architectures with LangGraph and Gemini.

What it demonstrates

• ReAct agent
• RAG agent
• Conversational agent with memory
• Document drafting agent
• Data-analysis agent with tool execution
• Supervisor-based multi-agent orchestration
• Specialized Coder · Enhancer · Researcher · Validator workers
• Conditional routing and shared graph state

Core stack

Python LangGraph LangChain Gemini RAG ReAct Multi-Agent Systems

Comparing custom CNNs and transfer-learning architectures for chest X-ray classification.

Models explored

Custom CNN · MobileNetV2 · ResNet50V2 · DenseNet-121

The repository reports its best experimental result with ResNet50V2 at 98% accuracy, alongside evaluation using precision, recall, F1-score, confusion matrices, and learning-curve visualizations.

LangGraph LangChain RAG ReAct Agentic AI Gemini OpenAI Embeddings Vector Search

Pandas NumPy Matplotlib Plotly Scikit-learn XGBoost SQL

FastAPI Django PostgreSQL Redis Celery RabbitMQ Docker Docker Compose Nginx Prometheus Grafana

                    ┌─────────────────────┐
                    │     Business Need   │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   System Design     │
                    │  APIs · State · Data│
                    └──────────┬──────────┘
                               │
              ┌────────────────┼────────────────┐
              ▼                ▼                ▼
        ┌───────────┐   ┌────────────┐   ┌─────────────┐
        │    RAG    │   │   Agents   │   │ Traditional │
        │ Retrieval │   │  Workflow  │   │     ML      │
        └─────┬─────┘   └──────┬─────┘   └──────┬──────┘
              │                │                │
              └────────────────┼────────────────┘
                               ▼
                    ┌─────────────────────┐
                    │  Backend Service    │
                    │ FastAPI · Workers   │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │ Deploy · Observe    │
                    │ Docker · Metrics    │
                    └─────────────────────┘

Not every project needs to be the headline. These repositories show the broader path that led to my current AI focus.

Agentic AI
    ├── Multi-Agent Orchestration
    ├── Tool Calling
    ├── Stateful Workflows
    ├── Memory & Context
    └── Evaluation / Validation

RAG
    ├── Document Ingestion
    ├── Chunking & Embeddings
    ├── Hybrid / Vector Retrieval
    └── Grounded Generation

AI Backend
    ├── FastAPI
    ├── Async Processing
    ├── PostgreSQL
    └── Distributed Workers

Production
    ├── Docker
    ├── Observability
    ├── CI/CD
    └── Cloud Deployment

I’m interested in working on Generative AI, Agentic AI, RAG, AI backend systems, and applied machine learning.