Level

Tags

the AI field has changed so much so curious how you're using AI now.

My stack:

search/daily driver (chatGPT). I switched from pplx to chatGPT bc o3 + web search is insanely good now

Coding. Split between windsurf/cursor. Cursor is just too expensive now with their recent changes to justify the price but its just so much better than windsurf. But I end up using windsurf mostly bc the price is the cheapest and their tab completions are unlimited, but its still kinda clunky and worse than cursor

Friends who have tried gemini CLI say its very good. I hear claude code is the gold standard right now but its incredibly expensive\
\
Cline works pretty well and im seeing good results with it

Perhaps people have already asked this question around but I could not find it. I want to know how are SWEs using charger or any other llms to up their game. I feel I am a bit behind in my team to make most out of it and want to start on this asap (don't want this to be the reason for my layoff :p). Regardless, I think this is changing the way we work, I work as a full stack and I aspire to be good at back-end. 

What are the skills/how can I use this technology to be a better dev?

On a side note, I have also been thinking about is this time to become a ai/ml developer? Or atleast a back end engineer who can scale these technologies?

How to use LLM for improving my work productivity

Senior Software Engineer at Taro Community

I am currently working as a PMTS in India Dev Center. I am best guy under my manager always get top ratings but I am introvert :) and not a good communicator. \
\
Next month leadership team ( VP and senior director) is visiting India and my manager has setup 1-2-1 with VP. \
\
What should I chat in 1-2-1. I am afraid if I dont have much to talk my VP will judge me.

Hi! I recently joined Meta as an E4 MLE ( from a DS background, not SWE). I am super-excited to be here and looking to ramp up and onboard as efficiently as possible. Would love advice from senior members who have gone through the process. My major concerns : 

1. I am the only E4 in my team and that worries me a bit.My TL designed an onboarding doc for me with tasks to familiarise myself with the processes and i have started with it (landed a couple diffs in my first week which ended this Friday). I am unsure how to pace myself here. Having heard of Meta’s pace, I am on a little worried of being left behind and going full throttle
2. How do I figure out impact in my team? Meta’s all about impact and I feel I need to figure out how to create it sooner than later.
3. I am trying to avoid the trap of getting lost in Meta’s Wikis. Whats the best way to understand the infra e2e tho?

MLE E4 : Advice for ramping up efficiently at Meta?

I'm a Sr. SDE working remotely from a different city than the rest of my team, which makes it hard to build relationships through informal chats with colleagues or my manager.

With constant reorgs, I believe there are new opportunities emerging, but I struggle with identifying them, and choosing the right project scopes that are truly impactful.

Could you please provide advice on:

1. How to surface meaningful project ideas in a shifting org?
2. How to get better at defining a project scope that aligns with org priorities and has visibility?

I want to improve in these areas so that I can contribute more strategically and grow in my role. Thank you!

How can I identify high-impact projects as a senior SDE?

Mid Level Engineer [L4] at Taro Community

I'm a mid-level engineer working on the design for a \~2 month long project. There are three other stakeholders involved in this (two L5's and an L6 who have been at the team for a while) and we've been discussing this project for the past \~3 weeks. My manager was on vacation and just came back.

Based on discussions over the past few weeks, I have a proposed solution. My manager (he has joined \~4 months ago) disagrees and has numerous concerns. So does another L6 who joined the team who he (presumably) asked to look at the design. They have a suggested approach that differs from mine.

The two other L5's and L6 on the team for a while generally agree with my solution and are unsupportive of the solution posed by my manager.

We had another discussion and left off with the conclusion that I am to speak with some engineers on core infra more familiar with the feasibility of my manager's solution. Ultimately, it's been decided that I will make the call on this by end of next week.

It's a significant, foundational project with large ambiguity and complexity, so the stakes are somewhat high.

What should I do to resolve this? I've added both L6's to the call with the infra engineers early next week. My assumption is that the infra team will say that alternative solution is not feasible.

Entry-Level Software Engineer at Unemployed

Hello, I made a project that started off as an MVP, but I feel like I vibed coded it and now it doesn't really work well as I hoped. But I see a learning to fix this AI code. How would you begin? I'm thinking of doing a refactor to try to decrease the file size and simply reduce the number of files and of course trying to follow React best practices and maybe introduce TypeScript to help.

Vibed coded my project - Now I wanna begin the refactor process to fix this

Some people say that debugging can take a few hours, others days depending on the code base (legacy or larger at a bigger tech company). On average, if I am at an earlier stage startup or a mid-size company (or big FAANGMULA), how long should this be? 

I feel like I've been wasting a lot more time on debugging for deploying (lots of minor syntax errors and stuff AI may not catch or careless mistakes and other dependency conflicts) than on actual real logic and I feel like it's taking forever. I just want to be sure I'm spending my time well for my projects, since I have so many, and even if this is one of my earlier ones, I'm worried that I'm spending too much time on it, even if my code quality is higher. 

So I started off in web development ages ago, and it has always frustrated me (open source frameworks being broken, debugging dependency from hell etc.) vs. mobile (which is a lot more straight forward and vertical rather than horizontal, and while slower with some companies being closed source like Apple, easier to deploy and not have to spend as much time debugging). AI is its own debugging hell since a lot of things were NOT working 10 years ago, but a lot has changed now. For the web devs/software engineers focused on this part of the stack, why do you still do it? I find that while open source web development is supposedly fast, it's so broken, and I'm finding more native development easier. With AI, it's like everything has gone rampant on crack though, so depending on where you are in the stack, it is all pretty relative. 

Thoughts and advice on where to focus on most? 

I've been through a lot of different parts of the stack and personally am so done with open source web development and it's not how I make my money mainly (it's more for branding purposes which is more than a fancy Jekyll or GH pages site). 

Sometimes stuff is taking forever to deploy. What do you do instead of looking at social media/email while your build is still building or deploying is taking forever (or even debugging)? How do you like to take breaks or feel like you're being productive rather than just sitting idly waiting?

What do you do during boring build times at work?

I’ve gotten feedback that I should ask questions earlier, and my manager mentioned that it’s always a tricky balance between working independently and looping others in. But in practice, I find this really hard to navigate.

For example, I try to be thoughtful about when I ask questions, especially when working with senior engineers. I’ll usually wait until I’ve tried multiple approaches and then time my questions for the morning, since responses can take a while. In the past, I’ve asked questions and ended up blocked all day waiting for a response, so I try to avoid that.

But the core issue is this: I often can’t tell when something is worth escalating. I don’t have a strong internal signal for when I should reach out to a senior engineer, a different team, or even a PM or manager. I genuinely don’t have know how to recognize when something is beyond me or when I should keep pushing on my own.

Any tips?

How do you get better at identifying when to escalate an issue?

Hey everyone,

\
I'm a mid-level engineer on a team where most of the others are seniors. I need to start reviewing their code, and I’m not sure what I should be focusing on.

When you do a code review, what exactly do you look for? Is it more about the code style—like variable names, function length, clarity—or should I also be evaluating technical decisions, like whether the overall approach makes sense?

It’s especially tricky because I’m still relatively new to the team, so I don’t have full context on a lot of the files or systems yet. I’ve seen advice about how to *write* code that’s easy to review, but less about how to *perform* a good review, especially when reviewing someone more experienced than you.

Would really appreciate hearing how others approach this!

What do you look at during a code review?

Entry-Level Software Engineer at Taro Community

Cluely is all the rage in Silicon Valley with the founder getting kicked out of Columbia and the company raising $15 million recently.

The thing is, if you visit <https://www.reddit.com/r/Cluely/>, you'll quickly see that most people think the product is just not very good. (Maybe there is a silent majority of people who think the product is great, but I wouldn't be sure of that.)

If the product is subpar and has no path to rocketship success, then what is the team doing with all of this funding, and what do investors think is possible with this idea? Are they more invested in the founding team (I will admit they are good at marketing and getting people's attention)? Even then, how are they different than all of the other LLM wrappers?

How do startups like Cluely raise so much money when their product is below average?

I am an ML Engineer with 6 years of experience at a mid-sized company, and I recently received an E5 MLE offer at Meta. I’ll be joining one of Meta’s recommendation systems teams, but my prior experience has been in a different ML field.

I’ve seen a lot of negative posts about Meta’s fast-paced and ruthless work culture on online forums, so I’d like to learn from other Metamates about what I can do to set myself up for success as an E5 MLE at Meta. Here are a few things I’ve planned so far:

**Before joining Meta**

1. Go through Taro’s onboarding materials -- [this onboarding course](https://www.jointaro.com/course/the-complete-onboarding-guide-for-software-engineers-succeeding-when-youre-new/), [onboarding playlist,](https://www.jointaro.com/playlist/9HLxktqQzqUDzukViJD8/)  [this course](https://www.jointaro.com/course/ask-great-questions-that-get-great-answers-quickly/), etc

2. Reach out to my Meta manager before my start date to understand the tech stack and projects planned for H2.

3. Ask my Meta manager or technical lead for any recommended external resources, such as books or online tutorials, that I should review before joining the team.

4. Read published research and blog posts about the recommendation platform I’ll be working on.

**After joining Meta**

1. Discuss expectations and the performance rubric with my manager, and start working toward these goals from day one.

2. Set up 1:1s with teammates and other cross-functional partners during my first 2–3 weeks.

3. Find mentors at Meta - both within my team and outside of it.

4. Volunteer for smaller tasks early on to build trust and learn the stack in low-risk contexts.

**I’m also especially interested in any Meta-specific advice on:**

1. Navigating calibration and performance review cycles to avoid surprises.
2. Identifying “impactful” work that matters at the E5 level.
3. Writing impact-oriented project updates that align with Meta’s culture.

In addition to what I’ve shared above, I’d really appreciate any tips, lessons learned, or suggestions for a successful onboarding and career at Meta as an E5 MLE!

New E5 MLE at Meta -- How Do I Set Myself Up for Success?

Software Engineering Intern at Taro Community

Hi Team,

I hope this message finds you well. I wanted to take a moment to introduce myself and seek your advice as I plan my next career move.

I currently have four years of experience as a Business Intelligence Engineer (BIE) at XXX. In this role, I’ve mainly worked with Python, maintained data engineering pipelines, and used tools like Amazon Redshift and Spark. I’ve also implemented simple machine learning algorithms—nothing too advanced like deep learning, but enough to get a solid grounding in the field.

Now, I’m looking to transition into a Software Development Engineer (SDE) role. With about six months to upskill, I’m both excited and a bit anxious—especially with the rapid rise of AI automation and generative AI agents in the industry

I’d really appreciate your perspective on a couple of options I’m considering:

- **Leveraging AI for Productivity:** With AI becoming more widely used, it seems essential for SDEs to be in the top 25% of their team by using AI tools to boost productivity. Is this something you see happening in your experience, and how can I best prepare for it?
- **Upskilling in Generative AI:** Since I have a basic understanding of ML, I was thinking of exploring generative AI. I’m considering courses like Udacity’s offerings and “Generative AI for All” from [DeepLearning.ai](http://DeepLearning.ai) on Coursera. Do you think this would be a valuable path for someone in my position?

Your insights would mean a lot as I navigate this transition. Thank you so much for your time and guidance.

How to transition from Business Intelligence Engineer to a Software Development Engineer working in AI?

Mid-Level Software Engineer at Taro Community

Hello,\
\
I started learning Gen AI end of 2023 and learned how to finetune an LLM from hackathons and by talking with researchers at the University while working nearby, similarly I learned using Pytorch and learned how to finetune AI models. \
\
I am a Gen AI expert, worked at a startup building ai tools from 0 to 1 and we got acquired in about half a year. I have a bachelors degree in computer science, only paper I wrote is during my bachelors didn't make much impact. \
\
I am specialized at AI Document Processing with Agents.\
For this I had to dive through all available OCR methods, document processing libraries available in python, java script and spend much time writing better ones. \
\
For a side project I am training an object detection AI Model , I saw a colleague who did a masters in AI train a model classic way with labeling using roboflow, and so I just copy it. \
\
These kinda jobs seem nice:

- <https://www.linkedin.com/jobs/view/4178264066/?alternateChannel=search&eBP=NON_CHARGEABLE_CHANNEL&refId=G7ZpsBh%2F6nT2L7BvUz8IoQ%3D%3D&trackingId=2okqutc4%2FnqMA0o%2BUYT%2Fhw%3D%3D>
- <https://www.linkedin.com/jobs/view/4120829208/?alternateChannel=search&eBP=NON_CHARGEABLE_CHANNEL&refId=VFEqSMSlyqZy%2B4YGl1zWGg%3D%3D&trackingId=1NxMwr1BYUuCrQNtOw9EHA%3D%3D>

How to approach getting this kind of role without masters degree/phd? From my position? Are there some others with these roles without the degree/phd? \
\
I'm wondering how I can leverage what I have now to get this kind of job. \
\
Since joining Taro I've gotten to know a lot about Meta from Alex and Rahul and seems like the nicest place to work from all the Tech companies out there. From the archetypes to culture, and managers being expected to code being in the trenches with the engineers if needed.

I wonder what approach to take to land a job at Meta, heard they're paying top buck for AI expertise from OpenAI guys, so they are willing to spend money on AI expertise that's interesting.

Getting job in AI Research Role without official Degree at Meta

I have my team matching round with director and senior manager for ic5 role tomorrow and the day after. What are some good questions I can ask them? Are there any courses for this?

Thanks

Mid-Level Software Engineer at Pharmacie centrale des hopitaux

What are some useful tactics to efficiently search for and evaluate whether a tool/library is right for the job, compare alternatives or decide to build a solution from scratch?\
\
What I find tricky is that differences between alternatives only become clear when the software becomes complex (which is often too late to go back) and that edge cases are unexpected and show up late once the tool is too entangled with your code.\
\
What I have tried is:

1. Ask AI (obvious), works well for suggesting alternatives, but pretty terrible at helping you pick the right one (not a big fan of relying on AI for things that can not be tested immediatly as it is really good at sounding knowledgeable even when it is not).
2. Architect the software as to make the choice reversible (if possible at all will still involve much rewriting)
3. I thought about trying to implement edge cases first but this assumes you already have exprience with whatever it is your are building.

I guess more broadly the question is how to approach decision making when testing alternatives is too expensive/time consuming and where prototyping is not very effective.

Choosing an off the shelf solution v. building from scratch

My code caused a recent incident, and I’ve just finished writing the postmortem. I’ll be presenting it to leadership for the first time and would really appreciate any advice on how to approach it effectively.

To give a bit of context:

- The API appeared healthy during the incident—latency, error rates, and general performance all looked normal.

- However, a boolean field in the API payload, implicitly relied on by the mobile app(IOS) not Android, was accidentally removed.

- During the code review, I tested on platform only and it was working okay and didn't test the other one.

- This dependency wasn’t documented in the codebase, and the field was being added by Component B, which enriches data from Component A.

- My changes were in Component A. Since Component B mocks Component A’s data in its unit tests, the missing field wasn’t caught in CI or during local testing.

- I fully acknowledge the mistake, and I want to be transparent and constructive in the review.

If you’ve got any tips—especially on how to communicate technical nuance and accountability clearly to leadership—I’d be grateful.

Thanks in advance!

Any advice on how to present an incident/SEV and its postmortem during an incident review with leadership?

Hey all. I’m still fairly new to Meta and Big Tech, and I’ve been feeling a bit conflicted about lunch breaks. My team usually eats together, and while they’ve been generally welcoming, I find that group lunches leave me feeling more drained than recharged. Especially with all the noise, the energy, and the language barrier (Non-english speaking Meta office).

I’ve been thinking about taking lunch alone just to have a bit of quiet time to reset, but I’m worried it might come off as antisocial or uncollaborative. Has anyone else navigated this? Is it okay to prioritize solo lunch for mental recharge, or will that reflect negatively?

Thanks for any thoughts.

Join The Discussion

Share insights on complex topics like system design, code review, tech leadership, interpersonal communication, and more.

What's your current AI stack?

How to use LLM for improving my work productivity

Topics of discussion 1-2-1 with VP