albusdd
@OfficialSamAyo
3x AI hackathon winner 🏆 | AI engineer | Full Stack Data Scientist | IoT | Speaker | sharing my knowledge to raise data legends | Building innovation everyday
✨ My AI Engineering Roadmap 🛣️ For those starting out as AI engineers/ML engineers, I have had a hard time giving all the answers to your questions. In this roadmap, I have highlighted the major topics you should learn. Pay attention to the footnote. A thread 🧵
yea, good observation.🫡 What i really want, is to see someone deduce what could actually be wrong from the training report. oya, try again.
Did you include early stopping ? From your train loss it's fluctuating The validation ACC is a bit far apart from the train as well Watch till it reaches 15 or 20 if there is no improvement you might need to stop
Sad to see @paperswithcode see the sunset. Thank you @huggingface for carrying the baton for the community. PWC is such an invaluable resource for the AI science community.
BREAKING: we've partnered with @metaai and @paperswithcode to build a successor to Papers with Code (which was sunsetted yesterday) PWC, founded by @rbstojnic and @rosstaylor90 has been an invaluable resource for AI scientists and engineers over the years (and an inspiration…
😂😂😂, My name's written all over the LAPO innovator. shoutout to @ConsonanceClub for this newsletter, this one got me laughing...
working with african dataset in computer vision is hard, lol Our intricacies in complex patterned dress and fabric will make an AI engineer say my God!😂. No wonder the western AI giants never really get our predictions right. #computervision #AIengineering #MachineLearning
To be there or not to be there? Me: I will be there🔥🔥🔥..How about you?
From Detection to Prevention: Unlocking Root Cause Analysis with Causal Inference in Manufacturing. This year, I'm coming in the Spirit and wave of the rigour-in-depth that you find in leaders such as: @cassiekozyrkov , @itszionking and the host. #oscafestgetready #oscafest25
From Detection to Prevention: Unlocking Root Cause Analysis with Causal Inference in Manufacturing. This year, I'm coming in the Spirit and wave of the rigour-in-depth that you find in leaders such as: @cassiekozyrkov , @itszionking and the host. #oscafestgetready #oscafest25
WE ARE BACK!✅ So, because I'm keen on building a community of AI builders on AWS in Africa☁️🟧 I will be sponsoring 7 tickets for builders interested in building AI solutions utilizing AWS. ⚡️Quote this with anything that suggests effort in this area!
EARLY BIRD TICKETS ARE LIVE! 🚨 Africa’s biggest data event is back and now’s your chance to lock in your spot at the best price! ⏳Early bird won’t last forever, grab yours now before prices go up! Secure your spot here: nestuge.com/DataFestAfrica… #DataFestAfrica2025 #DFA25
Join me tomorrow guys 🔥🔥. Register via: bit.ly/bir_agents @BlackInRobotics
Lastweek, As an AI/ML engr, I started designing an experiment for intent classification in convulated natural language conversations. Think of how Jarvis(of IronMan) can switch frm conversation to task execution(agentic) without loosing context of the conversation. #AI 🚀✨🤖💫🔥
A ranker is in and mostly a language model itself. This opens up alot of possibilities for what you can build as AI middlewares or mini semantic engines using high-performance rerankers like jina, cohere, mixedbread etc.
Curious about the AI Summer of Code? 🤔 Get all your questions answered at our AISOC Info Session this Saturday. Tune in on our Twitter Space as members of our organizing team @manlikesiq @OfficialSamAyo @itsDonmonc @_danielakhabue, walk you through everything you need to know.
Loving the responses so far — some sharp eyes in the mix! ML engineers lurking in the community, come out and flex your debugging skills. Can you spot what’s wrong in this ML training log snapshot? #MachineLearning #MLEngineering #DebuggingChallenge
Interesting phenomenon.🤔 The val accuracy dropped from 55.3% → 29.3% in 6 epochs Training accuracy stalled at ~50% then dropped to 31.2%. Then suddenly, we see the val accuracy take a u-turn and begin to progress. What can you deduce is happeing here? #MachineLearning #AI
Interesting phenomenon.🤔 The val accuracy dropped from 55.3% → 29.3% in 6 epochs Training accuracy stalled at ~50% then dropped to 31.2%. Then suddenly, we see the val accuracy take a u-turn and begin to progress. What can you deduce is happeing here? #MachineLearning #AI
good, good, good. but what is really wrong? I know it's speculative but be abit precise. You're almost there
The model stopped learning after the 4th epoch which means the model is not learning well at all. Hence something is wrong with either the architecture, the training loop or something even deeper than that.
This is all the information you should need. The idea is to observe the log, particularly the behaviour of the acc and val_acc and val_loss and predict what could be wrong. Oya, go back to the drawing board.
@OfficialSamAyo Can you provide more details about the other epochs?
good but that doesn't still identify what could be wrong. give me the ML in you, bro😄
The learning rate also changed at epoch 7, which suggests there's probably something designed to adjust (lower in this case) the learning rate as the model metrics nosedive The change in LR goes with the metric recovery
High quality research will still always standout despite the flooding of poor quality papers from half-baked researchers.
## The case for more ambition i wrote about how AI researchers should ask bigger and simpler questions, and publish fewer papers: