Jay.dot

AI Engineer | Football |

Building ML/DL models, APIs & Containers, Production deployment |

💼 Open to opportunities

[⚡️🇧🇼 @Priceless_MCI]

WHATEVER HAPPENS MANCHESTER CITY TILL I DIE 🩵

[Jibril Abdullahi Bala, PhD @DrJibrilBala]

I built an interactive scenario model for the 2027 Nigerian presidential race — Tinubu, Atiku, Obi — with adjustable parameters for incumbency effect, candidate-base retention across party moves, and opposition coalition scenarios.

[Marques Brownlee @MKBHD]

Google Lens was so ahead of its time, sheesh

[Jay.dot @Jerremiah__I]

@wadong3 @Real1_balogun I still see the structure cause a lot of the players that were brought in weren't even peps traditional kind of players. Imo it was more of a preparation for the next manager which was suspected to be Tuchel, but maresca still have a similar play type.

[Jay.dot @Jerremiah__I]

@amosuibk @IamMajet Since little tokens are flying around 😄
I’ll really appreciate mine as a referral for a ML engineer role
CV & portfolio are good to go, ready when needed 🙌

[Matt Dancho (Business Science) @mdancho84]

🚨 BREAKING: Microsoft launches a free Python library that converts ANY document to Markdown

Introducing Markitdown. Let me explain. 🧵

[Darlington Anyanwu | The Samurai Recruiter @official_cumi]

Normalize reaching out to potential employers/founders/co-founders on platforms that require them to lower their guard a bit.

You only have one chance. Do your research and shoot your shot.

Sell yourself like you have never and watch you go past tons of CVs

[spidey @lochan_twt]

ai intern interview question

[• nanou • @NanouuSymeon]

🐳 DOCKER COMMANDS CHEAT SHEET

[Jay.dot @Jerremiah__I]

@samfresh444 Big ups bro 💪

[Jay.dot @Jerremiah__I]

@soham_nayak04 @X AI/ML engineer, let's connect

[Jay.dot @Jerremiah__I]

@TheNewPaul1 @prospercode A guide😂
The intersection are developers

[Jay.dot @Jerremiah__I]

@prospercode The researching you do to find a solution to the problem is the real job description of a dev.

Syntax is just the delivery method

[Jay.dot @Jerremiah__I]

@AnantaVichara Why aren't you using a notebook?

Also if you don't like using a notebook you could run an interactive jupyter cell on a selected part of code inside the .py file. It's also fun to use for testing

[nocturnal @nocturnmachine]

I carry it all.
Not because I’m strong— but because I had to.

[Jay.dot @Jerremiah__I]

@nogo_vibes 😂 Plot twist, almost every "new" app is just an existing one with some new features. Facebook was MySpace with better features. WhatsApp was SMS with internet.

Execution > originality.

[Jay.dot @Jerremiah__I]

Just built an introvert vs extrovert prediction model!

Learned: hyperparameter tuning, threshold optimization, sklearn pipelines, FastAPI + Pydantic, and Docker containerization.

From zero to deployed in 3 days!

#MachineLearning #Docker #Python

[Marine Caous @marine_caous]

📍 Spotted in Paris today @huggingface - what would you do?

[clem 🤗 @ClementDelangue]

Everyone is talking about how we need more AI data centers (especially the ones who would mostly benefit from them) but why is no one talking about on-device AI?

Running AI on your device:
- Free
- Faster & takes advantage of existing hardware
- 100% privacy and control (you don’t send your data to an API)

[Hatman 🎩 @hatman]

Multimodal with just fine-tuning? ByteDance makes it happen with VoRA.

The new Vision as LoRA (VoRA) paper introduces a bold, streamlined approach to building multimodal LLMs—no vision encoder, no connector, no architecture bloat. Just LoRA.

Instead of bolting a vision tower onto an LLM, VoRA injects vision understanding directly into the LLM using Low-Rank Adaptation (LoRA). That means:

- No new inference overhead- LoRA layers are merged into the LLM after training.
- Frozen base LLM-Only LoRA + visual embeddings (~6M params) are trained, preserving language ability and ensuring stability.
- Image inputs at native resolution - No resizing, no tiling hacks—VoRA leverages the LLM’s flexible token handling.
- Bidirectional attention for vision-Instead of using causal masks across all tokens, VoRA allows vision tokens to attend freely—boosting context modeling.

To teach the LLM visual features:
- VoRA uses block-wise distillation from a pretrained ViT, aligning intermediate hidden states across layers. This improves visual alignment while keeping the LLM’s core untouched.
- The training objective combines distillation loss (cosine similarity between ViT + LLM visual features) and standard language modeling loss over image-caption pairs.

What does this get you?
- A modality-agnostic architecture ready for extension to audio, point clouds, and beyond.

This might be one of the most efficient takes yet on vision-language modeling. Excited to see how this evolves.