10 Things You Must Do As An Engineer Before 2026 Ends

The last 30% of the year is enough to change how you interview, if you're willing to be uncomfortable.

Author

Vishakha Sadhwani
September 15, 2026

September is here, which means roughly 70% of 2026 is already behind us. The remaining 30% is enough time to change how you walk into your next interview. But it won't happen by watching more tutorials.

These are ten things that feel uncomfortable on purpose. Each one puts you somewhere your knowledge gaps, communication gaps, or engineering gaps have nowhere to hide. Every resource below is free.

Before You Start: The Projects This Article Keeps Coming Back To

Several of the steps below ask you to read docs, reverse engineer something, point an AI agent at a codebase, or tell a story about a project. Instead of hunting for a different project each time, use these:

And two open-source AIOps projects worth studying:

  • K8sGPT — scans Kubernetes clusters and explains problems in plain English. CNCF Sandbox project since December 2023. Docs

  • HolmesGPT — an open-source AI agent for investigating production incidents, pulling from Prometheus, Loki, Tempo, ArgoCD, and more. CNCF Sandbox project since October 2025. CNCF write-up

1. Make a Presentation About Yourself

Why it's uncomfortable: Most engineers can explain a system for twenty minutes but freeze at "tell me about yourself." Putting yourself on five slides forces you to decide what actually matters.

How to do it: Build five slides and nothing more.

  1. Who you are — one sentence, no job-title soup.

  2. Your skills — only the ones you'd be happy to be quizzed on.

  3. What you've built — two or three projects, each with the problem it solved.

  4. What you enjoy doing — the kind of work that doesn't feel like work.

  5. Your dream role — specific. "Platform engineer working on Kubernetes at scale," not "something in cloud."

You get two things out of this. You're writing down where you want to go, which makes it far easier to move toward it. And you've just built your elevator pitch for interviews, networking, and introductions. Read slides 1, 3, and 5 out loud and that's your 60-second answer.

Free resources

Use AI the right way here: Write the slides yourself first. Then paste them into an AI assistant and ask it to play a skeptical hiring manager who asks three follow-up questions. Don't let it write the slides for you.

2. Read an Open-Source Project's Documentation Three Times

Why it's uncomfortable: No AI summary, no YouTube explainer. Just you and the docs. The first read will feel like a waste of time. It isn't.

How to do it:

  • Read 1 — Don't stop. Read start to finish. You won't understand most of it. Write down every term you don't know, but don't look anything up yet.

  • Read 2 — What does it do? Look up your list of terms, then read again. Run the quickstart. By the end, you should be able to explain what the project does in two sentences.

  • Read 3 — Why is it built this way? Now read the architecture and design pages. Look for the trade-offs: why this storage model, why pull instead of push, why a CLI instead of a service. This is the read that makes you sound senior in interviews.

Pick a project by level

Free guides for reading (and contributing to) open source

YouTube videos

Tip: Watch the videos after read 2, not before. Otherwise the video does the thinking for you, and that's exactly what this exercise is trying to avoid.

3. Find Five People in Your Dream Role

Why it's uncomfortable: Messaging strangers feels awkward, and silence feels like rejection. It's neither. People are busy.

How to do it:

  1. Search LinkedIn for your exact dream role (use the one from slide 5 above).

  2. Pick five people and study their career paths. What did they do two jobs ago? Which certifications, projects, or pivots show up repeatedly?

  3. Send each person one specific question. Not "can you mentor me?" and not "any tips?"

  4. No reply? Find five more. Keep going until you get one.

What a specific question looks like:

"Hi Priya, I noticed you moved from a support engineer role to SRE in about 18 months. I'm in support now and working through the CKA. Was there one skill or project that made the SRE interviews click for you?"

It's short, shows you did your homework, and can be answered in two lines.

Free places to find people who actually reply

4. Turn One Failure Into a Story You Can Tell Naturally

Why it's uncomfortable: You have to pick something that didn't work. Most people only rehearse wins, and it shows.

How to do it: Pick one college incident or work experience that went wrong. Put it into this shape:

  • What happened? The context, in two or three sentences.

  • What did you decide? The choice you made and why.

  • The good and the bad. What worked, what didn't, and what it cost.

  • What would you do differently? This is the part interviewers actually remember.

Say it out loud until it stops sounding memorized. Five or six run-throughs is usually where it starts to sound like a conversation.

Free resources

Practice partner: Use any AI assistant's voice mode. Tell your story, then ask it to respond like an interviewer and dig into the weakest part. It's free, available at 2am, and doesn't get bored on the sixth attempt.

5. Remove One Technology From Your Resume

Why it's uncomfortable: It feels like making your resume weaker. It actually makes it stronger.

How to do it: Go through every technology on your resume and ask:

  • Can I explain how I actually used it, not how it works in general?

  • Can I describe one thing that broke and how I fixed it?

  • Can I say why we used it instead of the alternative?

If you can't answer all three confidently, take it off. An interviewer will pick the one tool you're unsure about, and the conversation will go badly from there.

Free resources

Want to keep it on instead? Earn it back. Build something small with it this month:

6. Reverse Engineer a Product You Use Every Day

Why it's uncomfortable: Engineering blogs are dense, and one read isn't enough. You'll need several passes before the data flow makes sense.

How to do it: Pick WhatsApp, Netflix, or Instagram. Read their engineering blogs repeatedly, and for each post, answer these on paper:

  • How does data flow? Draw it. Client → edge → service → storage.

  • What was the bottleneck? Every good engineering post exists because something broke or got slow.

  • What's the caching strategy? What's cached, where, and for how long?

  • How is load balanced? Across regions, across services, across data centers.

After five or six posts, try drawing the whole system from memory.

Blogs / docs

Free system design references (keep these open while you read)

YouTube videos

Build something: Once you understand one system, build a tiny version of one piece of it. 5 cloud projects to build has good starting points.

7. Set Up an AI Coding Agent and Give It One Job

Why it's uncomfortable: Handing a codebase to an agent means reviewing work you didn't write, in a workflow you don't know yet.

How to do it: Install a coding agent on your laptop and give it one simple task:

  • "Review this codebase and tell me what it does."

  • "Fix this bug."

  • "Add a health check endpoint."

That's it. The goal isn't a perfect result. The goal is getting comfortable with how agents read code, ask for permission, make changes, and occasionally get things wrong.

Good first codebase: clone the repo from the Docker CI/CD project or terraform-zero-to-hero, and ask the agent to explain the pipeline or the Terraform modules.

Pick an agent

Free courses

Blogs / docs

YouTube videos

8. Give Your AI Agent One Custom Skill

Why it's uncomfortable: You have to write instructions clear enough for an agent to follow, which quickly shows you how clearly you actually understand the task.

How to do it: A skill is a folder with a SKILL.md file: instructions plus a short description that tells the agent when to use it. Skills follow an open standard (agentskills.io), so they work across multiple agent tools, not just one.

  1. Create a new skill, or install one from an open-source repo.

  2. Run a task without the skill. Save the output.

  3. Run the same task with the skill. Save the output.

  4. Compare them.

Example: a Dockerfile review skill

In Claude Code, project skills live in .claude/skills/ and personal skills in ~/.claude/skills/. Create .claude/skills/dockerfile-review/SKILL.md: 

---
name: dockerfile-review
description: Review Dockerfiles for image size, security, and build caching problems. Use when asked to review, fix, or optimize a Dockerfile.
---

# Dockerfile review

Check the Dockerfile against this list, in order. Report findings as a table with columns: issue, line, suggested fix.

1. Base image is pinned to a specific tag, never `latest`.
2. Multi-stage build is used if there is a compile or build step.
3. Dependencies are installed before `COPY . .` so layers cache.
4. The container runs as a non-root `USER`.
5. A `.dockerignore` exists and excludes `.git`, `node_modules`, and secrets.
6. A `HEALTHCHECK` is defined.

Do not rewrite the file unless asked. Findings first.

Now ask the agent to "review the Dockerfile" in the Docker CI/CD project repo, once without the skill and once with it. Look for what changed: did it check all six items, did the output format stay consistent, and did it hold back from rewriting the file?

Free courses

Blogs / docs

Open-source skills to install

  • anthropics/skills — Anthropic's public skills repo, including skill-creator, which helps you build new skills

9. Pick a Certification Learning Path

Why it's uncomfortable: Picking a path means committing to one, and paying for an exam makes it real.

How to do it:

  1. Find three certifications related to your dream career.

  2. Find the free and paid courses for each.

  3. Read the reviews.

  4. Read the official exam guide and curriculum.

  5. Sign up. Booking a date is what turns "someday" into a study plan.

Here's a starting menu for AI, Cloud, and DevOps, with official pages and free prep for each.

Cloud

AWS Certified Cloud Practitioner (or any practitioner-level cloud cert)

AWS Certified Developer – Associate / Azure Developer Associate (AZ-204)

Cloud roadmaps (crash courses): AWS · GCP · Azure

DevOps

RHCSA (Red Hat Certified System Administrator)

CKA (Certified Kubernetes Administrator)

CKS (Certified Kubernetes Security Specialist) — requires an active CKA

HashiCorp Terraform Associate

AWS DevOps Engineer – Professional / Google Professional Cloud DevOps Engineer

GitHub Foundations

DevOps crash courses: DevOps concepts · CI/CD · GitOps

AI

Claude Certified Architect – Foundations

Cloud AI certs (pick the cloud you already work in)

NVIDIA certifications (Generative AI LLM, Agentic AI, AI Infrastructure & Operations)

How to get the exam cheaper

Never book at list price without checking these first.

  • KSUG.AI — maintains a community discount code for Linux Foundation certs (CKA, CKS, CKAD, KCNA, KCSA). Check the site for the current code before you buy.

  • Microsoft Certification Voucher Offers (GitHub) — community tracker of live Microsoft vouchers and discounts.

  • Microsoft Virtual Training Days — free events that often include a free fundamentals exam voucher.

  • Google Cloud Innovators — free program with monthly lab credits; Google also gives out vouchers through Study Jams and GDG events.

  • AWS — passing an AWS exam has typically earned a 50% discount on your next one. Check your certification account after each pass.

  • NVIDIA — certification exams are frequently free for in-person GTC attendees.

  • Students — the GitHub Student Developer Pack includes a free GitHub Foundations exam voucher.

10. Share a Technical Project as a Story (and Watch Yourself)

Why it's uncomfortable: This is the most critical one on the list, and the hardest. Watching a recording of yourself explain something is genuinely painful. It's also the fastest way to find out what you don't really understand.

How to do it:

  1. Pick a project. If you don't have one, build the Docker CI/CD project or part of the End-to-End DevOps + AIOps project first.

  2. Start with: "Once upon a time, I was working with X, and then Y happened."

  3. Record yourself for 20 minutes. No script, no retakes.

  4. Watch the whole thing back.

What to look for when you watch it:

  • Where did you say "basically" or "it just works"? That's a gap in your understanding.

  • Where did you skip a step because it was hard to explain? That's what an interviewer will ask about.

  • Where did you lose the story and start listing tools? Bring it back to the problem.

  • Could someone who wasn't there follow the order of events?

Then record it again next week.

Free tools

  • OBS Studio — free, open-source screen and webcam recording

  • Your phone camera is completely fine too

Story prompts from the projects above

  • Docker CI/CD: "Once upon a time, I was deploying my app by hand, and then I pushed a broken build to production..."

  • AIOps: "Once upon a time, I was drowning in alerts from my Kubernetes cluster, and then I tried giving an LLM access to my logs..."

  • Terraform: "Once upon a time, I built my AWS network by clicking through the console, and then I had to rebuild it in a second region..."

Crash Course Quick List

Every video in one place, in case you just want the links:

The Whole List, in One Place

  1. Make a five-slide presentation about yourself.

  2. Read one open-source project's docs three times, with no AI summary.

  3. Find five people in your dream role and ask each one specific question.

  4. Turn one failure into a story you can tell naturally.

  5. Remove one technology from your resume.

  6. Reverse engineer a product you use every day through its engineering blog.

  7. Set up an AI coding agent and give it one job.

  8. Give that agent one custom skill and compare the results.

  9. Pick three certifications and sign up for one.

  10. Record a 20-minute project story and watch it back.

The point isn't to be uncomfortable for its own sake. It's to put yourself in situations where your knowledge gaps, communication gaps, and engineering gaps become impossible to ignore, while there's still time left in 2026 to close them.

All resources above are free courses, official docs, free labs, open-source projects, or free YouTube content. Tool pricing, free tiers, certification details, and discount codes change often, so confirm on the official pages before you sign up or book an exam. If a video link has moved, searching its title will usually find the current version.