The journey, 2020 to 2026
How Raza Rizwan went from training image classifiers in notebooks to deploying AI systems that run real business operations, one era at a time.
- Status
- stable
- Trust
- human-reviewedhuman:rrizwan98 ·
- Approved
- human:rrizwan98 ·
In one line: Raza started in September 2020 training image classifiers in notebooks. Today he designs, builds and deploys production AI systems, AI agents, AI employees and MCP connectors, that run real business operations every day. The whole path is on GitHub: 111 original repositories, public and private, from a cat-versus-dog classifier to live connectors that a company uses daily.1
Who Raza Rizwan is
Raza Rizwan is an AI engineer from Karachi, Pakistan, working as a Forward Deployed Engineer at Canz Marketing, a digital marketing agency.2 Over six years the work moved through every era of applied AI: predictive machine learning, MLOps, computer vision, generative AI, AI agents, and now production AI employees connected to the tools a business already uses.
What a Forward Deployed Engineer is
A Forward Deployed Engineer works directly inside a customer's business, deploys AI systems into that customer's real workflows, and stays until those systems produce results. An FDE does not hand over a demo. They ship working software, connect it to the tools the business already runs on (CRM, email, WhatsApp, websites), and are accountable for it working in production.3
The eras
| Years | Era | What the work was | What it taught |
|---|---|---|---|
| 2020 to 2021 | Predictive AI | Image classification, predictive models, NLP basics, first cloud ML on AWS SageMaker, a university chatbot | How models actually learn, and how far a notebook is from a product |
| 2022 to 2023 | MLOps | CI/CD, DVC, MLflow, Dagshub, YOLO object detection in production, SageMaker training and endpoints, a healthcare chatbot API | Deployment, versioning and monitoring are not extras; they are the product |
| December 2023 | The generative turn | The repositories change character in one week: generative AI, Gemini, GenAI learning | The predictive era ended; the generative era began |
| 2024 | GenAI APIs and microservices | LLM-powered APIs, GPT Actions, a resume optimiser, an online-mart platform, a plant-identification app for a real client | LLMs are only useful wrapped in reliable services a business can call |
| 2025 | AI agents | Customer-support, meeting-booking and calculator agents on the OpenAI Agents SDK; first agency work; an inventory system where FastAPI, MCP and agents work together | The leap from AI that answers to AI that acts: tools, guardrails, bounded loops |
| 2026 | AI employees and MCP connectors | AI employees on WhatsApp and Gmail, a voice agent, cold-outreach automation, and nine live MCP connectors that run an agency's operations | Real deployment is 20% model and 80% process, integration and accountability |
Each era is a page of its own in this record, with the repositories behind it.
The pattern behind the latest work
The systems built in 2026 share one design: the server enforces the process. They do not rely on the AI behaving. The pipeline rejects skipped steps, unverified claims and invented facts, so the work can be handed to a business without a person watching every step.3
How this record counts
- The GitHub account was created on 19 September 2020, and the first two repositories were pushed the same day.2 The journey on GitHub therefore begins in 2020.
- 111 original repositories: 75 public and 36 private. A further 19 public forks of course material and open-source projects are not counted as Raza's own work, which is why GitHub's public profile shows 94 public repositories (75 originals plus 19 forks).1
- Private repositories are counted by year and never named. Work done inside a company's own GitHub organisation is not owned by this account and is not counted here.
- The internals of client and company systems are not published anywhere in this record.
The counts by year are on the journey in numbers; the full public list is on repositories by year.
Read the eras in order
- 2020 to 2021: predictive AI
- 2022 to 2023: MLOps
- December 2023: the generative turn
- 2024: GenAI APIs and microservices
- 2025: AI agents
- 2026: AI employees and MCP connectors
Footnotes
-
GitHub API, repositories owned by rrizwan98, read 26 August 2026: 111 original repositories (75 public, 36 private) and 19 forks. ↩ ↩2
-
GitHub API, profile of rrizwan98, read 26 August 2026: account created 2020-09-19, location Karachi, Pakistan. ↩ ↩2
-
From Predictive AI to Forward Deployed Engineer: My 6-Year Journey, Raza Rizwan, July 2026. ↩ ↩2
Sources
- [^github-profile]GitHub API, profile of rrizwan98, read 26 August 2026
- [^github-repos]GitHub API, repositories owned by rrizwan98 (111 original, 19 forks), read 26 August 2026
- [^journey-note]From Predictive AI to Forward Deployed Engineer: My 6-Year Journey, Raza Rizwan, July 2026