Build Log #01: Why I'm Building Analog Data and What Almost Stopped Me

Six months ago I launched the first Analog Data cohort with 12 students. I wasn't sure it would work. I'd been teaching embedded systems workshops for companies — Bosch, Siemens, a few defence PSUs — but this was different. Consumer-facing. Online. A cohort model I'd never run before.

Here's the honest version of what happened.

The Problem I Kept Seeing

After five years of corporate training, I noticed a pattern. Engineers who'd worked 2–5 years in the industry couldn't answer basic production questions:

• "How do you size a FreeRTOS task stack?"
• "What happens when the WiFi reconnects while your MQTT task is publishing?"
• "How do you handle a watchdog reset in the field without losing data?"

These aren't advanced questions. They're the questions that come up on your second production deployment. But nobody teaches them. YouTube channels show you how to blink LEDs with ESP32 Arduino. University labs use 8-bit AVRs. Corporate onboarding assumes you already know.

The gap between "completed B.Tech ECE" and "can ship production embedded firmware" is enormous — and nobody was bridging it systematically.

What I Almost Got Wrong

My initial instinct was to build a recorded course. 60 videos, self-paced, sell it on Udemy. That's the obvious move.

Three things stopped me:

1. Recorded courses have 90% dropout. I've seen the data from platforms I've worked with. The completion rate for self-paced technical courses is genuinely 5–10%. The people who finish are already motivated; the people who need structure get lost by week 2.

2. Embedded systems needs hardware in the loop. You can't learn FreeRTOS stack sizing by watching a video. You have to hit a stack overflow, see the backtrace, fix it, and understand why. That requires a live environment, real feedback, and someone to explain what the crash dump means.

3. The bottleneck isn't content — it's debugging. Every student who reaches out to me has the same problem: they're stuck on a crash they can't understand. The video answered the happy path. The production failure is never in the video.

So I built a cohort model instead. 12–15 students, 8 weeks, live sessions twice a week, code review on every assignment, WhatsApp group for real-time debugging help.

What the First Cohort Data Showed

After 8 weeks with 12 students:

• 10/12 completed all assignments (vs 5–10% on recorded platforms)
• 9/12 received salary increases or made a role transition within 3 months
• Average reported skill jump: from "I can do Arduino projects" to "I can write FreeRTOS firmware and connect to AWS IoT Core"
• Most valuable sessions (by feedback score): Stack overflow debugging (week 4), OTA update implementation (week 6), MQTT reconnect logic (week 5)

The number that mattered most: 9 of 12 students reported that they used something from the cohort in their actual job within the first month after finishing.

What's Coming in Cohort 2

Based on feedback, I'm adding:

1New in Cohort 2:
2├── Week 0: C fundamentals bootcamp (for students who need it)
3├── Week 5: ESP32 + AWS IoT Core full production setup (expanded from 1 to 2 sessions)
4├── Week 7: Edge AI intro — TFLite Micro on ESP32-S3
5└── New: Monthly alumni sessions (debugging war stories from the field)

The curriculum is still 8 weeks, but tighter. I cut two sessions on protocol theory that nobody used and replaced them with more debugging practice.

The Part Nobody Tells You About Building an Edtech

The hardest part isn't the curriculum. It's the marketing.

I'm an engineer. Writing LinkedIn posts about "why embedded systems is the future" feels unnatural. Running paid ads feels expensive without data on what converts. Cold outreach feels spammy.

What actually worked: writing technical content like this and letting people find it. The top-performing posts on my LinkedIn aren't "enrol in my cohort" — they're "here's the actual cause of your ESP32 watchdog reset." Engineers share technical content. They don't share ads.

This blog is part of that strategy. If you're learning something from what I write here, you're the kind of person Analog Data is built for.

Where to Find Analog Data

If you're an embedded engineer in India who wants to make the jump from hobby/service-level to production firmware:

• Website: analogdata.in
• Cohort 2: Applications open — DM me on LinkedIn with your current experience level
• Free resource: The ESP32 Production Firmware series on this blog is the curriculum preview

No sales pitch. If it sounds like what you need, it is. If it doesn't, the free content is here anyway.