DataDetective101

Data Detective | Statistics graduate turning raw data into insights with AI & automations | Building AI in public | Kenya-focused applications

[DataDetective101 @Datadetective10]

@StanleyMasinde_ This is like telling James Orengo does know the Kenyan Constitution😂😂 diabolical

[DataDetective101 @Datadetective10]

@_iamhabeeb I can even smell this image

[DataDetective101 @Datadetective10]

@WilliamsRuto Bitching🏹😂

[DataDetective101 @Datadetective10]

@claudeai

Hey team, this issue keeps happening on my account: tokens are being consumed with no output generated.

Is there a way to recover the wasted tokens? Any fix or workaround would be appreciated! 🙏

[DataDetective101 @Datadetective10]

@CallMeDoct Why lie

[DataDetective101 @Datadetective10]

LLMs are powerful, but the real engineering skill is knowing how to build around them.

Choose the right model.
Understand its limits.
Test the output.
Design the workflow properly.
That is the path from “using AI” to building with AI.

[DataDetective101 @Datadetective10]

Frontier LLMs are impressive.

They can synthesize information, write structured answers, expand rough ideas, and help with coding.

But they are not perfect.

They can miss recent events, use old APIs, hallucinate, or sound confident while being wrong.

Evaluation matters.

[DataDetective101 @Datadetective10]

Today I compared frontier model families like GPT, Claude, Gemini, Grok, and DeepSeek.

One lesson stood out:

A good AI engineer should not depend blindly on one model.

You need to understand strengths, weaknesses, cost, speed, reasoning ability, and use case fit.

[DataDetective101 @Datadetective10]

Chat models are strong for conversation, summaries, writing, and interactive use cases.

Reasoning models are stronger for difficult problem-solving.

Base models are important when you want to fine-tune toward a specific skill.

Different models, different jobs.

[DataDetective101 @Datadetective10]

Day 3 of my LLM Engineering journey is complete.

Today I learned more about frontier LLMs and how they differ:

Base models
Chat/Instruct models
Reasoning/Thinking models
The deeper I go, the more I realize LLM Engineering is about choosing the right model for the right task.

[DataDetective101 @Datadetective10]

Today I worked with the OpenAI Python client.

Simple lesson:

The client is basically a cleaner way to call LLM APIs from Python instead of manually handling raw HTTP requests.

Small concept, but very important if you want to build real LLM applications.

[DataDetective101 @Datadetective10]

Day 2 of my LLM Engineering journey is done.

Today I learned about closed-source vs open-source LLMs.

Closed-source: GPT, Claude, Gemini
Open-source: Llama, Mistral, Qwen, Gemma, Phi, DeepSeek, GPT-OSS

The LLM ecosystem is bigger than just ChatGPT.

[DataDetective101 @Datadetective10]

Day 1 reflection:

Starting this LLM Engineering course feels like the right step.

With my Statistics background, I want to position myself around AI + data + automation.

The goal is to become someone who can build useful AI systems, not just talk about AI.

Day 2 tomorrow.

[DataDetective101 @Datadetective10]

My focus for the next 8 weeks:

Learn by building.

Not just watching tutorials.
Not just collecting certificates.
Not just reading about AI.
I want to build real LLM apps, RAG systems, AI agents, and portfolio projects that solve practical problems.

[DataDetective101 @Datadetective10]

Today I started setting up the foundation for my LLM Engineering.

The course covers Hugging Face, LangChain, RAG, QLoRA, agents, and a final autonomous AI project.

I’m especially interested in building AI tools that work with data, documents, reports, and business workflows.

[DataDetective101 @Datadetective10]

One thing I’m learning early:

LLM Engineering is not just about prompting.
It’s about building useful systems around models:
data
documents
retrieval
tools
agents
workflows
That’s the direction I’m taking as a Statistics graduate moving into AI engineering.

[DataDetective101 @Datadetective10]

Day 1 of my LLM Engineering journey is complete.

Today I started the AI Engineer Core Track: LLM Engineering, RAG, QLoRA, Agents course.

My goal is simple:

Move from just knowing Statistics to building real AI systems with LLMs, RAG, agents, and automation.

[DataDetective101 @Datadetective10]

Finished Day 1 of LLM Engineering.

Set up the repo, Cursor, uv, .env, OpenAI API key, and made my first OpenAI API call from scratch.

Also started building a website summarizer.

Small start, but this is exactly the kind of hands-on learning I wanted.

[DataDetective101 @Datadetective10]

I’ve started my 8-week LLM Engineering journey.

Course: AI Engineer Core Track — LLM Engineering, RAG, QLoRA, Agents
Background: Statistics
Goal: AI Data & LLM Engineer
I’ll be building with RAG, agents, LangChain, Hugging Face, QLoRA, and Python.
Time to build.

[DataDetective101 @Datadetective10]

I just started my LLM Engineering journey.

I’m taking AI Engineer Core Track: LLM Engineering, RAG, QLoRA, Agents and documenting the process publicly.

Background: BSc Statistics
Goal: AI Data & LLM Engineer
Focus: RAG, agents, automation, real portfolio projects.