AI

Bonnet brand workflow illustration

Bonnet AI

Bonnet is an AI brand-development product built around a clear delivery flow. A user starts with a creative brief, then the system pushes that work through research, strategy, creative direction, moodboards, and exportable outputs. The interesting part is not chat. It is the orchestration needed to turn a messy brief into something a team can actually review and ship. Tech Stack: Next.js, React, Django, Channels, PostgreSQL, OpenRouter, Supabase Source: Private (commercial product) ...

January 15, 2025 · 3 min · Muhammad Hassan Raza
Obelisk AI content system illustration

Obelisk

Obelisk is an AI-powered marketing platform that helps teams create content with brand consistency. It orchestrates specialized AI agents for SEO, email marketing, brand voice analysis, and strategy, all within a multi-tenant SaaS architecture with space-level isolation. Tech Stack: FastAPI, LangGraph, PostgreSQL, Vertex AI, Redis, Google Cloud Source: Private (commercial product) · Book a call to discuss My role: Team product at Entropy Labs. I led product and backend work around agent orchestration, retrieval, checkpointing, guardrails, and multi-tenant runtime behavior. ...

January 20, 2025 · 16 min · Muhammad Hassan Raza
Anatomia callback workflow illustration

Anatomia Healthcare

Anatomia is a care workflow product centered on callbacks and case review. The core loop is not a general healthcare record system. It is a nurse-facing workflow where calls, transcripts, AI triage, patient context, follow-up work, and escalations all have to move cleanly between staff roles. Tech Stack: FastAPI, React, PostgreSQL, AWS Cognito, AWS S3, AWS KMS, Redis, OpenAI, Vapi Source: Private (healthcare client) My role: Team project. My work was concentrated on the backend workflow layer, transcript and triage handling, case-state plumbing, and the follow-up paths that had to stay tied to the right patient record. ...

January 10, 2025 · 2 min · Muhammad Hassan Raza
LLMs and User Experience

LLMs Can't Save Bad UX

I wrote about AI features users actually want a while back. The TLDR was: stop building chatbots, start building smart defaults. That post got shared, people agreed, and then most of them went back to building chatbots. This is the sequel. The one about what happens when AI features ship and nobody uses them. When the LLM is working correctly and the product is still failing. When the problem was never the model. ...

March 18, 2026 · 9 min · Muhammad Hassan Raza
Structured logging cover

Structured Logging for AI Debugging

When an AI coding assistant tries to help you debug a production issue, it reads your logs. If your logs are scattered console.log calls with inconsistent formatting, the AI can’t help you. It doesn’t know which log lines belong to the same request, what the timing was, or what the error context means. Evlog is a structured logging library by Hugo Richard, designed around the “wide event” pattern. One structured event per request, with all context attached. I’ve been using it in my projects and it’s particularly useful when you’re debugging with AI tools, because the log output is machine-readable by design. ...

March 5, 2026 · 4 min · Muhammad Hassan Raza
Legal intake fraud scoring cover

Scoring Fraud in Legal Intake Calls

I worked on RISQ, a fraud detection system for legal intake calls. Mass tort law firms receive thousands of calls from potential claimants, and a significant percentage are fraudulent: coached callers reading from scripts, people who never actually used the drug or product in question, or repeat callers under different names. RISQ listens to these calls, scores them for authenticity, and recommends whether to transfer the caller to a closer, flag them for review, or quarantine the call. ...

February 20, 2026 · 5 min · Muhammad Hassan Raza
Multi-agent middleware illustration

Multi-Agent LLM Middleware: Lessons from Marketing Accelerant

Marketing Accelerant is an AI-powered marketing analytics platform I worked on at Entropy Labs. It runs 15+ specialized LLM agents for Brand Voice, Creative Content, CMO Strategy, SEO, Email Campaigns, Google Ads, Meta Ads, Video Studio, and more, all serving enterprise clients through a single FastAPI backend. The agents themselves aren’t the hard part. The hard part is everything around them: model selection, context management, cost control, error recovery, and human approval. This post covers the middleware architecture that makes it work in production. ...

January 10, 2026 · 7 min · Muhammad Hassan Raza
Model Context Protocol diagram

Model Context Protocol: Why This Matters More Than You Think

Every few months, something gets released that looks like infrastructure plumbing but turns out to matter more than the flashy launches. Model Context Protocol (MCP) is one of those things. If you’re a developer working with LLMs, MCP will change how you integrate AI into your workflows. Here’s an early-adopter perspective on what it is, why it matters, and how to actually use it. What Problem Does MCP Solve? Today’s AI tools are context-starved. You paste code into ChatGPT, upload files to Claude, manually copy database schemas into prompts. Every session starts from scratch. Every context window is a blank slate. ...

November 15, 2025 · 7 min · Muhammad Hassan Raza
Extended thinking interaction model diagram

Extended Thinking in LLMs: A Mental Model for Developers

Extended thinking isn’t just “model thinks longer”—it’s a fundamentally different interaction model. If you’re prompting extended thinking models (Claude Opus, o1) the same way you prompt standard models, you’re leaving most of the value on the table. This post is a developer’s mental model for working with these systems: when to use them, how to prompt them, and what trade-offs to expect. How Extended Thinking Actually Works Standard LLMs generate tokens one at a time, each token conditioned on everything before it. The model “thinks” only as fast as it speaks. Ask it to solve a complex problem, and it often commits to an approach in the first few tokens, then rationalizes that approach even if it’s wrong. ...

September 25, 2025 · 7 min · Muhammad Hassan Raza
AI features users actually want diagram

AI Features Your Users Actually Want (Hint: Not Another Chatbot)

The graveyard of failed AI features is full of chatbots nobody asked for. Every product team I talk to has the same story: leadership watched a GPT demo, got excited, and mandated “we need AI in the product.” Three months later, there’s a chatbot in the corner of the app that 3% of users have tried and 0.5% use regularly. As CIO at Entropy Labs, I’ve been on both sides of this. I’ve built AI features that users loved and killed features that seemed brilliant in demos but died in production. Here’s what I’ve learned about the difference. ...

August 10, 2025 · 6 min · Muhammad Hassan Raza