My Journey Into AI, Systems Engineering & Product Thinking

October 15, 2025 (1mo ago)

Introduction

In the last few months, I've been working across multiple domains — AI engineering, full-stack development, embedded systems, cloud automation, and even academic program management. This journey wasn't just technical; it taught me how to learn fast, how to build systems that scale, and how to overcome unknowns one step at a time.

This blog summarises how I approached complex problems, the technical stacks I explored, and the mindset shifts that helped me transform scattered tasks into an engineering process.

🧩 1. Starting Point: "I don't know everything, but I can learn anything."

The first shift was accepting that no project comes with complete clarity. Whether it was:

...every task demanded learning on the go, Googling, testing, and adjusting.

The mindset that helped:

🔑 Learning Rule 1 — Do > Read

I learned to set up AWS, Cloud9, Next.js, Redis, Docker, Python environments by actually doing it — not waiting to feel "ready."

🔑 Learning Rule 2 — Break things early

Most progress came from:

Each break gave a better understanding of the system architecture.

⚙️ 2. Engineering the AI Backend: From Simple Queries to Production RAG

At Innovites, I worked on building a 1M+ data RAG architecture.

This pushed me into:

🧠 How I approached it

Step 1 — Start with the smallest working unit

I built a tiny RAG demo with 50 documents.

Step 2 — Experiment aggressively

Tried different embedding models, memory optimization, chunking techniques.

Step 3 — Scale with guardrails

Improved context windowing, caching, and query rewriting.

Step 4 — Production mindset

Monitoring, logging, user intent detection — not just "make it work."

🏆 Result

I became more confident in designing AI workflows from scratch — not copy-pasting code from GitHub but solving business problems with AI.

🛰️ 3. ESP32–CAM Object Detection: From "It's Not Working" to "I Understand Why"

Running a camera module at high altitude was new territory.

Challenges hit immediately:

But this is where engineering starts.

🔨 How I overcame it

1. Understood the constraints

ESP32-CAM has 520 KB SRAM, so image processing must be minimal.

2. Offloaded detection to Roboflow API

Fastest outcome → API-based inference.

3. Designed a parallel Edge AI workflow

For long-term Jetson deployment.

4. Created a camera cycle strategy

A reliable 2-second capture loop.

🎯 Learning

You can't fight hardware limitations. You design around them.

💻 4. Full-Stack Architecture: Turning Features Into Systems

From student dashboards to SaaS credit systems, I built multiple app flows.

🔧 What I learned technically

🔨 My engineering approach

  1. Identify the "core entity" (User, Team, Report)
  2. Then build schemas
  3. Then API routes
  4. Then UI flows
  5. Then automation

🌱 Learning

Frontend is easy. System design is the real skill.

🧭 5. Managing Teams & Projects: From Solo Builder to Project Lead

Working with 28 students and 6 teams taught me:

💡 What I applied

1. Break the entire project into "high-level workflows".

For example:

2. Assign roles based on skill

Not interest alone.

3. Documentation > Verbal instructions

Clarity removes 70% of mistakes.

🧠 6. The Meta-Skill: Learning How to Learn

Across everything, one skill stood out:

Rapid learning with structured experiments

Whenever I hit an unknown:

🔍 I ask myself:

This systematic learning helped me navigate:

🌟 Conclusion: The Journey is a System, Not an Accident

If I had to summarize:

And the combination of these made me not just "a developer," but an engineer who can learn anything on demand.

LinkedIn