Anas Sayed

AI Engineer building Decision Intelligence & Agentic AI. Specialty: FinTech Risk Engines, RAG & LLM Agents. 🏦 🤖 | FastAPI • XGBoost • Docker • CrewAI

[Anas Sayed @AnasBuilds_AI]

🧠CNNs don’t see images—they see features 🔍
🔹 Conv2D → finds edges & spots
🔹 MaxPooling → compresses data
🔹 Dense → final classification Hardest part?

Capturing signal, not noise.
📈 Model: 93% accuracy | Next: Data Augmentation
👉 https://t.co/lD2N4THtIn
#AI #CNN

[Anas Sayed @AnasBuilds_AI]

📁 Code & White Paper:
https://t.co/EFQfEnOxyr
🌐 Demo: https://t.co/dy7cL47dVB

[Anas Sayed @AnasBuilds_AI]

🏦 I don’t just train models—I build decision engines.

Just launched a Production‑Grade Credit Risk Engine:
⚡ FastAPI + Docker
🌲 XGBoost + SHAP (XAI)
📊 Regime Detector + FinBERT Sentiment Transparent, scalable, regulator‑ready AI.

#AI #FinTech #MLOps

[Anas Sayed @AnasBuilds_AI]

Portfolio Link: https://t.co/GfA3gg6b7f

[Anas Sayed @AnasBuilds_AI]

ML Core Milestone [100% COMPLETE]
🏁 6 weeks → 2 Projects → 1 Mindset shift.

🏎️ Car Price Prediction | 📉 Customer Churn. Built for reliability, not “perfect” scores

⚡️ Roadmap: Past✅ Math+ML | Present🧠 Deep Learning | Future🤖 LLMs.

Link: (in comments)
#AI #ML #Python

[Anas Sayed @AnasBuilds_AI]

📉 AI in Finance ≠ just stock prediction.

Built an Enterprise Credit Risk Engine 🏦:
✅ XGBoost for high recall
✅ SHAP for explainability
✅ FastAPI + Docker for scale
✅ Macro‑layer with NIFTY & rates

Transparency > Accuracy. 🚀

#AI #FinTech

[Anas Sayed @AnasBuilds_AI]

📉 Accuracy ≠ Business Value.

A churn dataset: 90 % stay, 10 % leave.
Predict “all stay” → 90 % accurate, yet misses every churner 💸.

Focus on Recall (catch leavers), Precision (avoid false alarms), & F1‑Score. Confusion Matrix is the real compass.

#AI #ML

[Anas Sayed @AnasBuilds_AI]

Heres other:
https://t.co/lD2N4THtIn

[Anas Sayed @AnasBuilds_AI]

Memorization ≠ Intelligence 🧠❌

99% train acc but 60% test? That’s not AI—it’s Ctrl+C, Ctrl+V.

Overfitting kills real-world ML.
✅ Train/Test splits
✅ Validation Loss
✅ Simpler models (Occam’s Razor)

Stability > Numbers ⚡️

#AI #ML #Python #BuildInPublic

[Anas Sayed @AnasBuilds_AI]

Linear Regression = “How much?” 💰 Logistic Regression = “Which one?” 🏷️
🏎️ Car Price App: R²=0.69
📉 Churn Predictor: Accuracy=81%
Takeaway: Confusion Matrix > Accuracy. Recall matters!
🧠⚡️ Don’t just import models—understand the math.
#AI #ML #BuildInPublic

[Anas Sayed @AnasBuilds_AI]

🛑 Stop obsessing over the Model.
🧼 Start obsessing over Preprocessing.

My ML accuracy jumped after:
1️⃣ Handling NA values
2️⃣ Removing Outliers
3️⃣ Feature Scaling

Respect the data → better predictions.
🗺 Progress: 65% | 📂Repo in comments
#AI #ML #Python #DataScience

[Anas Sayed @AnasBuilds_AI]

Deep Learning Portfolio: https://t.co/lD2N4THtIn

[Anas Sayed @AnasBuilds_AI]

Data Analysis = rearview.
Machine Learning = windshield 🏎️💨.

Wrapped Model Building milestone:
🔹 Features > Models
🔹 Clean Data > Algorithms
📂 Projects: Churn & Car Pricing
🗺 Status: ML Core ✅

#AI #ML #Python #BuildInPublic

[Anas Sayed @AnasBuilds_AI]

My Portolio: https://t.co/GfA3gg6b7f

[Anas Sayed @AnasBuilds_AI]

📈 Stats course ✅
Realized every AI “engineering” problem = math problem.

Cosine similarity → geometry
Confidence scores → probability
Agents → logic optimization

AI engineering = math + code.
Now, back to building 🚀

#MathForAI #BuildInPublic

[Anas Sayed @AnasBuilds_AI]

Stop celebrating Accuracy on imbalanced datasets.🛑

In Churn Prediction, "Accuracy" is a vanity metric.

• Accuracy= How often you are right.
• Recall= How often you catch the target.

If you miss the churn, 99% accuracy is useless.

I optimized for Recall.📉
#DataScience #AI

[Anas Sayed @AnasBuilds_AI]

Unpopular opinion:

No need to master 100% of SQL.
Just master the 20% that cleans data for models.

• GROUP BY + HAVING → Feature Engineering
• LEFT JOIN → Data Enrichment
• CASE WHEN → Categorical Encoding

If you can’t query it, you can’t model it. 📊
#SQL #AI

[Anas Sayed @AnasBuilds_AI]

@Seannywilson Email

[Anas Sayed @AnasBuilds_AI]

@crewAIInc Github: https://t.co/CCkNd8hX5Y

[Anas Sayed @AnasBuilds_AI]

🚀 Built my first Multi‑Agent System with @crewAIInc

4 agents running a FinTech marketing agency:
🎯 Strategy | 🔍 Research | 🎥 Scripts | ✍️ Blogs

No human in the loop.
Agentic workflows > chatbots.

Code in comments 👇
#AI #Python #CrewAI #BuildInPublic