Mehedi
Hasan

Brain Tumor Detection System

AI-powered medical imaging with 95%+ accuracy

A deep learning-powered medical image analysis system for detecting and classifying brain tumors from MRI scans. Uses state-of-the-art CNN architecture with advanced data augmentation and transfer learning.

AI-Powered E-Commerce Platform

Smart shopping with ML recommendation engine

Full-stack e-commerce solution featuring an AI recommendation engine, real-time inventory management, Stripe payment integration, and an admin analytics dashboard.

Smart Library Management System

Digital library with automated fine calculation

Comprehensive digital library management with book tracking, member management, automated fine calculations, barcode scanning, and an analytics dashboard.

Multilingual Sentiment Analysis API

Real-time NLP inference across 12 languages

Production REST API for real-time sentiment analysis and text classification using fine-tuned transformer models, supporting 12 languages with sub-100ms inference.

LLM Chatbot with RAG Pipeline

Context-aware AI assistant with custom knowledge base

Intelligent chatbot using Retrieval-Augmented Generation with a custom vector knowledge base, multi-turn memory, and streaming responses.

Visual Workflow Automation Platform

No-code automation builder with 50+ integrations

Drag-and-drop workflow automation platform with a visual builder, scheduled tasks, webhook triggers, and real-time execution monitoring.

Architecting Multi-Agent Systems in Production: Supervisor vs Peer-to-Peer Orchestration with LangGraph

Single LLM agents struggle with context window pollution and hallucination cascades when tasks exceed 5-10 reasoning steps. Here is how we architect production multi-agent systems using supervisor-worker patterns, typed shared state, and deterministic cyclic graphs.

Beyond Naive RAG: Building Production Hybrid Retrieval with Dense Vectors, BM25, and Cross-Encoder Rerankers

Pure vector search fails when queries contain exact product SKUs, code identifiers, or acronyms. We dive deep into Reciprocal Rank Fusion (RRF), Cross-Encoder reranking, and chunking heuristics that achieve 96%+ context precision.

Building Fail-Safe Tool-Enabled Agents: Pydantic Schema Enforcement, Sandboxed Execution, and Error Recovery

LLM tool calling is fundamentally probabilistic. When an agent emits malformed arguments or calls destructive APIs in a blind loop, systems fail. Here is our blueprint for deterministic schema validation, self-correcting retries, and isolation.

Provider-Agnostic LLM Architecture: High-Availability Routing, Fallbacks, and Rate-Limit Resiliency

Hardcoding vendor SDKs creates single points of failure and vendor lock-in. We explain how to architect a production LLM gateway featuring automated model fallback, latency-based load balancing, and unified token telemetry.

Engineering AI Evaluation Frameworks: Continuous Benchmarking with LLM-as-a-Judge and Ragas in CI/CD

How do you know if your prompt update, chunking tweak, or model upgrade actually improved your AI application? We examine how to build CI/CD evaluation pipelines using Ragas, TruLens, and deterministic rubric judges.

Production Guardrails: Defending Against Indirect Prompt Injection, Jailbreaks, and PII Leaks

When agents read external websites, parse customer emails, or ingest user uploads, they become vulnerable to indirect prompt injection. We outline a defense-in-depth security architecture that protects data integrity without degrading model utility.