Plot. Not many of us are privileged to understand or grasp reality as is. We all perceive the world as per our understanding; your growth depends on that level. Lack of emotional intelligence is such a red flag — if human interactions become less interactive compared to AI, then what’s the point?
My ChatGPT has more personality than an actual person. I think more than 80% of the population is in autopilot mode. No awareness, no opinions, no moral compass — just BS random spiral thoughts. Such a waste.
Twist. In my daydreaming world, I want to buy a palace in Italy, convert it into a boutique hotel (for creative satisfaction), live a free life, have an art studio, bike to beaches, pet count (double digits obvio), open a Michelin-star fusion restaurant which specializes in weird flavors & menus, fly to countries as an artist. But guess how far I am from this? Very far. But I will reach there — I have made up my mind.
Sometimes your struggles aren’t you — not your abilities, not your flaws — but the people with whom you decided to share your life. It could be family, friends, etc.
You have to onboard empty-minded, stubborn, autopilot individuals ‘cuz solo rides aren’t allowed.
It’s a crippling disease.
It takes a real companion to build a life.
But to live? You can live with a rock.
Twisted Reality. My daydream might be a reality to many, but my reality is that it’s a dream for me. In my current reality, I am responsible for building data solutions with AI for a bioscience company. & I see a lot of promising innovations on the biotech level.
If you do the most hated things long enough, you start liking it — such as how I started to embrace the uncertainty of my career, as I love programming & problem-solving. But I have to onboard 50+ companies, hundred rounds of interviews, and throw in a few bunch of degrees. Don’t forget the constant visa filings and bare minimum benefits. On top of that, maintain a happy mrg life? Which already has its own sprint cycle. Looking at the future ahead, society also requires me to have a kid(S) ASAP.
But today, I am here to share a hope. How AI is helping us see the world.
AI Revolutionizes Eye Health: From Personalized Treatment to Vision Accessibility
Artificial intelligence (AI) is rapidly transforming the field of ophthalmology, enhancing how clinicians detect, manage, and treat eye diseases. At recent global research meetings — including those hosted by the Association for Research in Vision and Ophthalmology (ARVO) — scientists presented groundbreaking work on how AI-powered algorithms are reshaping both clinical and public health strategies.
AI in Risk Assessment and Disease Progression
Researchers are now using machine learning models trained on thousands of retinal images and patient histories to detect early biomarkers for diseases such as diabetic retinopathy, age-related macular degeneration (AMD), and glaucoma. For example, DeepMind’s AI system, co-developed with Moorfields Eye Hospital in the UK, demonstrated human-level performance in diagnosing over 50 eye conditions using 3D OCT scans — helping prioritize urgent cases and personalize treatment plans.
A recent 2024 study published in Nature Biomedical Engineering described a deep learning model that can predict the 5-year progression of AMD with over 85% accuracy, outperforming traditional clinical predictors. These models evaluate risk based on a combination of retinal thickness, lesion patterns, genetic markers, and lifestyle data.
Enhanced Imaging and Diagnostic Precision
AI tools like Google Health’s automated retinal screening system can now process and analyze high-resolution images to detect microaneurysms and hemorrhages in diabetic retinopathy in under a minute. This enables earlier intervention, particularly in underserved areas where access to ophthalmologists is limited.
Similarly, algorithms such as REFINED-OCULUS, developed in 2023, use unsupervised learning to map subtle changes in optic nerve structure, allowing ophthalmologists to monitor glaucoma progression with unprecedented sensitivity.
Empowering Vision-Impaired Individuals
Beyond diagnostics, AI is powering assistive technologies to support people with vision loss. Apps like Be My Eyes and Seeing AI use real-time computer vision and natural language processing to describe surroundings, read text, and recognize faces. A newer project, Open Access NavAI, presented at ARVO 2025, showcased a wearable device that combines LiDAR, GPS, and AI scene understanding to provide intuitive spatial cues for blind users navigating urban spaces.
These tools not only enhance mobility but are also being integrated with smart home systems, offering greater independence and safety for individuals with low vision.
Public Health and Population-Level Insights
By aggregating anonymized eye imaging data across populations, AI systems are also being used to detect disease trends. For example, an AI model developed by Stanford researchers tracked the rise of myopia in schoolchildren across East Asia and recommended policy interventions for screen time and outdoor activity — a real-world example of AI informing public health strategy.
Here’s a simple but realistic Python example using an AI algorithm (deep learning) to classify retinal OCT images to detect eye disease (e.g., age-related macular degeneration vs. healthy retina), using a convolutional neural network (CNN).
We’ll use the popular UCSD Retina OCT dataset — feel free to download and adjust paths if running locally.
AI Algorithm for OCT Image Classification (AMD vs. Normal)
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import os
# Set image size and paths
img_height, img_width = 128, 128
batch_size = 32
data_dir = "/path/to/OCT2017" # Update this path
# Preprocessing and data augmentation
train_datagen = ImageDataGenerator(
rescale=1./255,
validation_split=0.2,
horizontal_flip=True,
zoom_range=0.2
)
train_data = train_datagen.flow_from_directory(
os.path.join(data_dir, "train"),
target_size=(img_height, img_width),
batch_size=batch_size,
class_mode='binary',
subset='training'
)
val_data = train_datagen.flow_from_directory(
os.path.join(data_dir, "train"),
target_size=(img_height, img_width),
batch_size=batch_size,
class_mode='binary',
subset='validation'
)
# Build a simple CNN model
model = Sequential([
Conv2D(32, (3, 3), activation='relu', input_shape=(img_height, img_width, 3)),
MaxPooling2D(pool_size=(2, 2)),
Conv2D(64, (3, 3), activation='relu'),
MaxPooling2D(pool_size=(2, 2)),
Flatten(),
Dense(128, activation='relu'),
Dropout(0.3),
Dense(1, activation='sigmoid') # Binary classification: AMD vs. Normal
])
# Compile the model
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Train the model
history = model.fit(
train_data,
validation_data=val_data,
epochs=10
)
# Save the model
model.save("oct_ai_model.h5")
Follow a tech philosopher’s perspective on AI and the world, while navigating an AI career.