Nurul Akter Towhid

  • Home
  • Nurul Akter Towhid

Nurul Akter Towhid I'm Nurul Akter Towhid - Artificial Intelligence & Data Science expert

29/08/2025

If you’re building an AI startup or product and looking for advice or even a small investment, feel free to reach out. I want to support early-stage founders — of course, only if the idea and product truly make sense to me.

📧 Reach me at: [email protected]

16/05/2025

Okay, is AI taking over everything? Maybe just the really boring bits first! 😄

Saw the news that Wikipedia is starting to use generative AI. But before we panic, it's not writing the articles (phew!).

Instead, they're using it to help their human editors with the time-consuming stuff:
- Background research
- Finding sources
- Translation tasks

Basically, AI is becoming the super-assistant, handling the grunt work so the amazing human volunteers can focus on quality, accuracy, and the stuff that needs real critical thinking.

It’s less "robot overlord" and more "really efficient helper".

Kind of refreshing to see AI being framed as a tool to boost human effort, not just replace it. Makes you think, right?

What's one tedious task in your work you'd happily hand over to an AI assistant? Let me know below! 👇

02/05/2025

Okay, who else scrolls through endless app reviews trying to figure out if that app is actually good? 🙋‍♂️

"Life-changing!" says one review.
"Total garbage!" screams the next.
My thumb gets tired just thinking about it.

Saw that Apple is now using LLMs to summarize App Store reviews. Basically, AI reads all the rants and raves so you don't have to, giving you a quick snapshot.

It's kinda like having that one friend who can perfectly summarize a 2-hour movie in 30 seconds. Super handy!

Feels like LLMs are quietly moving out of the "big scary AI" headlines and into genuinely useful little corners of our digital lives. Less Skynet, more slightly-smarter-toaster vibes, maybe? 😉

It's interesting seeing these complex models become practical tools, even if its just helping us decide whether to hit 'download'.

What's the most surprisingly normal place you've seen AI pop up lately?

26/04/2025

Is Bangladesh going to write its own AI story, or just read the one written by others?

Reading Zunaid Kazi's recent article in Dhaka Tribune really got me thinking. That feeling of potential slipping away if we don't act together... it's a bit worrying, right?

He makes some solid points about shaping our AI future:

* We can't keep working in our own little bubbles. A national council bringing government, academics, industry, and even diaspora experts together? Makes so much sense.
* Getting AI-savvy isn't just about learning to code. We need lawyers, doctors, journalists, farmers – everyone – understanding how it impacts their world.
* Let's build an environment where our own AI startups have a real shot – access to funds, data, and places to test their ideas.
* And maybe most importantly: Make AI work for everyone. From city professionals to rural communities, in Bangla and regional dialects too.

It’s kind of like having all the ingredients for an amazing dish, but needing a shared kitchen and a recipe everyone agrees on before we can cook! 🧑‍🍳

The article asks if we'll move with purpose or risk falling behind again.

What do you think is the biggest hurdle we need to tackle first to make sure AI truly benefits all of Bangladesh? Curious to hear your thoughts!

25/04/2025

Heard about the CRM that basically runs itself? 🤔 Two founders from Bangladesh just raised $2.6M to build it.

We've all been there, right? Drowning in CRM updates instead of actually talking to customers. That feeling when you spend more time logging activities than *doing* activities... it's real!

Saw this news about Octolane AI and it caught my eye. Founded by two guys who experienced the pain of traditional CRMs firsthand, they're building what they call a "self-driving" AI CRM.

What does that mean?

* AI automatically pulls details from emails, meetings, docs to update records.
* It analyzes interactions and suggests next steps.
* It can even generate personalised follow-up emails.

Basically, it aims to kill the manual data entry beast.

Think of it like having a super-smart assistant who *only* does the CRM admin you secretly (or not so secretly) hate.

As someone fascinated by practical AI applications, this is cool stuff. It’s not just adding AI features *to* a CRM, it’s rethinking the workflow *with* AI at the core. Less admin burden, more time for actual human connection in sales – sounds like a win.

Plus, seeing founders tackle such a huge, established market ($100B+!) is prtty inspiring.

Makes you wonder... what's the most time-consuming admin task AI could solve for *you*? Curious to hear your thoughts!

21/04/2025

Okay, is the hiring world turning into bots interviewing bots now? 🤖

Just read about AI-generated fake applicants flooding tech companies. Using AI to fake IDs, resumes, even interview answers!

It's a bit ironic, isn't it? Considering how many "ghost jobs" seem to float around. Maybe the fake applicants are just trying to find the fake jobs? 😉

Seriously though, it adds another layer of complexity for genuine folks trying to land a role and for recruiters trying to find them. Makes you think about trust and verification in this digital age.

How do we keep the 'human' in Human Resources amidst all this tech? Curious to hear your experiences! 👇

20/04/2025

My brain is still buffering... OpenAI just dropped GPT-4.1 with a 1 million token context window?! 🤯

Remember wrestling with context limits? Trying to squeeze huge documents or long conversations into prompts sometimes felt like trying to fit an elephant into a matchbox.

Well, going from GPT-4o's 128k to 1 million tokens feels like upgrading from a pamphlet to an entire library. Mind = blown.

So yeah, OpenAI launched GPT-4.1 last week. Key takeaways:

Massive 1M token context (Seriously, think of the possibilities!)
Reportedly better & faster performance (especially for coding tasks)
Cheaper to run than GPT-4o (My wallet appreciates this)
Plus new 'Mini' and 'Nano' versions for different needs.

Looks like they're doubling down on refining the current generation while we wait a bit longer for the much-hyped GPT-5. Honestly, given how complex these things are, taking time to get it right makes sense.

It feels like the pace just doesn't slow down in this field!

What's the first thing you'd do with a 1 million token context window? Or are you more excited about the potential of the smaller, faster models? Curious to hear your thoughts!

20/04/2025

Ever feel like your data went 10 rounds with a blender? 😅

Data cleaning... definitely not the most glamorous part of our job, right?

Hours can vanish just trying to fix inconsistencies and fill gaps. It's easy to feel like you're just tidying up instead of doing the 'cool' AI stuff.

But today reminded me:
➡️ Sometimes the biggest insights hide in the messiest data.
➡️ Don't just clean it, understand it. The 'errors' might tell their own story.

It's a bit like being a detective, sifting through clues. The real gold isn't always shiny and perfect. ✨ It often requires patience and a closer look.

What surprising discoveries have you stumbled upon while cleaning up data chaos? Curious to hear your stories!

18/04/2025

Ever have one of those weeks at work? 😅

You know the feeling... a project doesn't land quite right, a presentation goes sideways, or that idea you loved gets a polite "no thanks."

It's easy to spiral into thinking, "Maybe this isn't for me."

But honestly, I think those moments are just part of the process. It stings, sure, but it's not the end game.

The real loss only happens if we stop trying altogether. Every stumble, every "nope," is just... data. Data for the next attempt, the next project, the next idea.

As long as we're learning and adjusting, we're still moving forward. It’s less about avoiding failure and more about getting really good at bouncing back. 💪

What helps you reframe a setback and keep pushing?

13/04/2025

Bangladesh just opened its doors to the stars (well, Starlink satellites anyway!). 🚀

Anyone else remember the struggle of finding reliable internet outside the big cities? Sometimes felt like sending data via carrier pigeon might be faster. 😉

So, seeing the news that Bangladesh granted Starlink a license got me thinking.

This could be huge for bridging the digital divide.

Imagine better connectivity for remote schools, businesses, maybe even smoother remote work for some of us data folks living further out?

It's like building a new digital highway, straight from orbit.

Of course, it's early days, and rollout takes time (and cost is a factor!), but the potential is definately exciting. It feels like another step towards a more connected Bangladesh.

What opportunities do you see this opening up for tech and data science in Bangladesh? Curious to hear your thoughts!

13/04/2025

Fighting fire with... algorithms? 🔥

It's fascinating to see Sharjah exploring AI to make buildings safer from fire hazards. Reading about the collaboration between the American University of Sharjah and the Civil Defence Authority really caught my eye.

They're not just talking theory; research is underway, aiming for tangible results possibly by year-end and implementation next year. Think AI predicting potential disaster spots or drones acting as first responders – like insights shared from a US fire chief.

It’s a great reminder that AI isn't just about chatbots or image generation; it has serious potential in public safety and infrastructure resilience. Plus, Sharjah is already tackling fire safety structurally, replacing building facades with fire-resistant materials. 🏗️

While tech like AI and robotics offers powerful tools, Dr. Rami Hawileh's point resonates – ultimately, individual awareness and responsibility are crucial in fire safety. Technology supports us, but doesn't replace us. 💡

What are your thoughts on using AI for predicting and preventing disasters like fires?

12/04/2025

Okay, let's talk LLM apps! 🎉 You build something amazing, test it out, and chef's kiss it works perfectly... for you.

But then comes the tricky part: showing your team (or even your friends) that it's actually good, not just "works on my machine" good. We've all been there, right?

Building these apps is getting easier with tools like Langchain and APIs from OpenAI/Gemini. But proving they're reliable? That’s a whole different puzzle. 🧩

Turns out, just running a few examples isn't enough. The context I read highlighted a few ways to really measure performance:
➡️ Standard benchmarks & metrics (like BLEU/ROUGE) give us numbers, but might miss the point for your specific app.
➡️ Human feedback is golden for nuance, but slow and can be subjective.
➡️ Even using another LLM to judge is an option (wild, right?), though it has its own quirks.

The big takeaway? Evaluation isn't one-size-fits-all. It needs good data that reflects real use cases (including the weird edge cases!) and metrics tied to what the app should achieve. It's about building trust and making sure our cool creations are genuinely useful.

How are you tackling LLM app evaluation? Always curious to hear what works for others!

Address


Website

Alerts

Be the first to know and let us send you an email when Nurul Akter Towhid posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Nurul Akter Towhid:

Shortcuts

  • Want your business to be the top-listed Media Company?

Share