Most systems die somewhere between the demo and production. AI features die there fastest.

I build the layer that keeps them alive.

7+ years of backend systems, AI agents, and data infrastructure, built where a silent error costs real money. I engineer for the failure case first, so nothing I ship fails quietly.

View projects

Name Md (Toufiq) Toufiquzzaman

Designation Senior Software Engineer

Experience 7+ years of professional experience

Focus AI Agents · MCP & Agent Tooling · Event-Driven Automation · Crawling & Data Pipelines · Production Reliability

Industries Automotive (Currently Working) · E-commerce · Real Estate · Music & Streaming · Fashion & Apparel · Electronics

Currently Building AI agents, backend, and data systems behind advertising and SaaS for 600+ dealerships, at a Canadian tech company.

Location Remote: Canadian time zones 4+ years (current), US time zones 2 years

About

The failures that never throw an exception

Reliability defaults

idempotent writes
reconciliation
retries with backpressure
drift detection

I've spent 7+ years building and operating production systems where reliability, correctness, and scale directly impact real businesses. For nearly five years, I've worked on the multi-tenant B2B platform behind 600+ automotive dealerships, building distributed crawling infrastructure, event-driven pipelines, server-side tracking, and data synchronization systems that keep inventory, campaigns, and analytics consistent across multiple platforms.

In systems like these, correctness is a chain. If a crawler falls behind, data goes stale. If data goes stale, downstream systems operate on the wrong state. And many data failures never throw an exception: they surface later as incorrect inventory, broken integrations, or wasted ad spend. That constraint taught me to engineer for failure first, and to build observability that detects drift before customers do.

More recently, I've been applying that same production discipline to AI. I built an AI Agent that now runs inside our internal systems to detect failed crawls and data drift in production, diagnose the underlying issue, repair the pipeline, and verify the result automatically. The loop that previously required human intervention can now close itself.

Everything I build starts with the same question: What happens when this fails, and who finds out first? That question has shaped how I build reliable systems for 600+ businesses, and now it's the same principle I bring to AI systems.

Personal Projects

Solo built & running in production

Two products of my own, designed, built, and operated alone outside my day job: first commit to on-call. ScrapeOra automates product data for store owners. AgentOray scores how ready a storefront is for the AI agents that will browse it.

ScrapeOra

Solo-built product automation SaaS · Live in production

scrapeora.com

AI AgentDetects breakage and repairs importers on its own

Self-healingImporters rebuild when a supplier changes layout

Always currentScheduled re-sync keeps price and stock in sync

Solo-builtDesigned, shipped, and operated end to end

Problem

Dropshippers build their catalog by hand. They open a supplier's site, copy the title, description, images, price, and variants, paste it into their own store, redo the keyword research, and publish. Then it goes stale: the supplier changes a price or runs out of stock, and every one of those listings has to be found and corrected manually.

Scrapers were the obvious fix, and they broke constantly. The moment a supplier changed its page structure, the importer stopped working and someone had to re-map it by hand.

What I built

An end-to-end SaaS: paste any supplier link and publish-ready products stream into Shopify, WooCommerce, BigCommerce, and more. Hybrid extraction backend, AI-written descriptions, idempotent upserts, and secure credential handling, so re-imports update listings instead of duplicating them.

Engineering impact

An AI agent watches the importers. When a supplier changes its layout and extraction breaks, the agent detects it, diagnoses what moved, and repairs the importer itself: no manual re-mapping, no support ticket. Scheduled re-sync keeps price and stock current from every supplier to every connected store. Built and operated end to end by one engineer.

AI Agent Self-Healing Automation AI Enrichment Web Crawling Data Pipeline Python SaaS

AgentOray

Solo-built agent-readiness scanner · Live in production

agentoray.com

Agent Readiness Score0–100, across 22 diagnostic checks

Agent-eye scanningReads a site the way an AI agent would

Prioritized fixesRanked by effort against points recovered

Solo-builtDesigned, shipped, and operated end to end

Problem

Shopping is moving to AI agents. People increasingly ask ChatGPT, Perplexity, or Gemini to find and compare products, and the agent reads storefronts in seconds without a human ever browsing them. Most stores were built for human eyes and search crawlers (heavy JavaScript, missing structured data, blocked paths), so the agent simply can't read them. The store isn't ranked badly; it's invisible.

What I built

A scanner that visits a store the way an agent does: robots.txt, sitemaps, product pages with JavaScript disabled, structured data, and policy pages. It scores the store 0–100 across 22 checks in five categories (access, structure, feeds, content, and policies), and returns a prioritized fix list with the effort and points recovered for each one.

Engineering impact

Every finding ships with the evidence behind it, so a fix can be verified rather than claimed. Built for the web that agents will browse, not the one search engines indexed.

AI Agents Agent Readiness Structured Data Web Crawling E-commerce Automation

How I can help

Designed, built, scaled, and operated on teams building backend systems and AI infrastructure.

Backend & Distributed Systems

Design, build, and operate distributed backends in Go, Rust, TypeScript (NestJS), and Python on PostgreSQL, Redis, Kafka, AWS, and Docker. Multi-tenant systems that stay consistent under load and keep serving when a dependency doesn't.

AI Agents & AI Infrastructure

Agents trusted with real infrastructure: I ship and operate a self-healing agent that repairs production crawl and data pipelines on its own. RAG pipelines, MCP tool design, schema-constrained outputs.

Web Crawling & Data Pipelines

Distributed crawling and event-driven pipelines moving millions of records and billions of events monthly: deduplicated, reconciled, and current, with an extraction layer built to survive source sites that change underneath it.

Production Reliability

Failure-case engineering for systems where a silent error costs real money: idempotent writes, reconciliation, retries with backpressure, and the kind of testing that assumes the dependency will be down.

System Architecture & Scalability

Architecture designed to be operated, not just shipped: multi-tenant isolation, event-driven boundaries, and hard integration surfaces (OAuth and token lifecycle, GA4 and Meta server-side tracking) that scale from the first tenant to the six-hundredth.

Observability, Cost Optimization & Performance

Per-request cost, latency, and success rate attributed to tenant and feature. Observability that catches drift before a customer does, and turns the next performance or spend decision into a measurement, not a guess.

Decision Policy for AI Autonomy

Which decisions an agent may make on its own, scored on reversibility, blast radius, explainability, and error cost. Permissions enforced at the infrastructure level, outside the prompt. Autonomy earned, not assumed.

Experience

A track record at production scale

Built across real engineering challenges: crawling infrastructure, data feeds, and systems that have to stay up.

10,000+

Cups of coffee

2,500+

Days in production

600+

Automotive dealerships

Billions

Data points processed: automation, self-healing systems, and production operations

Nov 2021 – Present

Senior Software Engineer

sMedia Ventures Inc.

Build and operate the multi-tenant B2B platform behind 600+ automotive dealership tenants: crawling and data pipelines handling millions of records and billions of events monthly, with server-side tracking across GA4 and Meta. Shipped a self-healing AI agent that now detects and repairs pipeline failures in production, alongside RAG and MCP-based integrations.

May 2021 – Oct 2021

Software Engineer

Data Assistant Corp.

Maintained systems for automotive dealership operations, handling system stability, data management, and ongoing support to keep business workflows running smoothly.

Jan 2020 – May 2021

Junior Software Engineer

CodeSurfer Ltd.

Started my career building and maintaining e-commerce systems, focused on stability, feature improvements, and consistent performance.

Trust layer

What people I've worked with say

Trust isn't claimed. It's earned through consistent delivery. A few words from clients and teammates.

Toufiq and I worked together for several years on the product team at sMedia and during that time I found him to be exactly the kind of teammate one would dream of working with. He always worked extremely diligently and the quality of his work was nothing less than exceptional. He was the kind of teammate you could always rely on to handle his responsibilities in a way that set the team up for success. One of Toufiq's best qualities was that he took initiative to identify problems and find solutions on his own, often without needing direction from his superiors or other teammates. He's extremely hardworking and given the opportunity I would definitely work with Toufiq again in the future. This recommendation was not written by AI.

Trevor Ash · Business Development Manager

Md is an excellent and exceptional developer with up-to-date skills needed for any small or big project. He delivered just as he promised. He's a good listener and always willing to help. We had a lot of bugs on our music site and needed to add extra functionality and increase the app's speed and performance — Md did just what we were looking for. I can recommend him to anyone at any time and will be eager to recruit him for my next project. Thank you for making our project come to life.

P. Dennis Sambola Jr · Founder

Toufiq is an extremely honest and highly communicative web and app developer. He helped us with our tasks, was very transparent, and always sought professional solutions. We highly recommend him for technical and complex tasks, including website development, app integrations, and app development.

Dimytrii Tupikin · Full-Stack Marketer

Contact

Let's build something solid

The most interesting problems sit where backend reliability meets AI infrastructure. That's the layer I'm building in.

Email toufiq.py@gmail.com

Location Bangladesh

LinkedIn in/runtimeranger GitHub @runtimeninja

Email me Back to top

Send a message

Tell me what you're building. I'll get back to you.