Writing archive
Posts
A calmer, searchable reading space for ideas, notes, and technical explainers.
Post
The 99% Accuracy Trap - Why High Scores Often Signal Failure in Machine Learning
High accuracy can hide overfitting, leakage, imbalance, and shortcut learning. Real ML quality is measured by generalization, robustness, calibration, and business reliability.
Post
RedSage - Building a Cybersecurity Generalist LLM with Agentic Data Augmentation
Tiny Recursive Models prove that thinking longer through recursion can outperform thinking bigger through scale in AI reasoning.
Post
Tiny Recursive Models - Why Recursion Beats Scale in AI Reasoning
Tiny Recursive Models prove that thinking longer through recursion can outperform thinking bigger through scale in AI reasoning.
Post
dataRLsec - Safety, Security, and Reliability With Robust Offline Reinforcement Learning for DPAs
Junk data and data poisoning both corrupt AI at training time by distorting learned representations, and robust offline reinforcement learning defends against this by identifying and down-weighting untrustworthy data before it can damage reasoning and behavior.
Post
Extreme Ultraviolet Lithography
EUV lithography is the near-impossible machine that turned atomic-scale physics into a manufacturable process, keeping Moore’s Law alive when everything else failed.
Post
T5 – Text-To-Text Transfer Transformer
T5 is a unified encoder-decoder Transformer architecture that frames every NLP task as a text-to-text problem, enabling exceptional performance across diverse domains.
Post
Activation Functions
Activation functions neural networks. They determine whether a neuron should be activated or not by transforming the input signal into an output signal.
Post
The Evolution of the Web: From Web1 to Web3
Explore the progression of the internet from Web1 to Web3, highlighting how user roles, technology, and ownership models have evolved over time.
Post
GPT – Generative Pre-trained Transformers
GPT is a decoder-only, autoregressive language model built on the Transformer architecture. It predicts the next token in a sequence and powers state-of-the-art natural language generation.
Post
BERT - Bidirectional Encoder Representations from Transformers
BERT is a transformer-based model that uses bidirectional self-attention for improved language understanding and task-specific fine-tuning.
Post
Transformers
Transformers are a deep learning architecture utilizing self-attention, cross-attention, and multi-head attention to efficiently process sequences, enabling powerful models for tasks like language modeling and transfer learning.
Post
Denoising Diffusion Probabilistic Models [Diffusion Models]
Denoising Diffusion Probabilistic Models (DDPMs) generate high-quality images by progressively adding and then removing noise, learning to reverse the process to create realistic images.
Post
Generative Adversarial Networks [GANs]
Generative Adversarial Networks (GANs) use two competing neural networks to generate realistic data, with applications in image generation, synthetic data creation, and more.
Post
REST APIs - Guide
RestAPIs are most widely used API framework and this is a breif description about what REST APIs are.
Post
Git and Github
Git is a powerful version control system that enables developers to track changes, collaborate efficiently, and manage project versions with ease.
Post
Prompt Engineering Frameworks
Craft clear prompts by defining roles, tasks, goals, and desired outcomes for better AI results.
Post
Enhancing Generative AI [GenAI] models: Fine Tuning vs. RAG
An overview of Fine Tuning vs. Retrieval Augmented Generation (RAG) techniques for enhancing Generative AI models, highlighting their strengths, limitations, and ideal use cases.
Post
Simulated Annealing [SA]
Simulated Annealing (SA) is an optimization algorithm inspired by the annealing process in metallurgy, known for its ability to escape local optima and find near-global solutions in complex problems.
Post
Ant Colony Optimization [ACO]
ACO is an optimization algorithm inspired by ant behavior, using pheromone trails and heuristic information to efficiently solve complex problems like routing and scheduling.
Post
Particle Swarm Optimization [PSO]
PSO is an effective, easy-to-implement optimization algorithm inspired by social behavior in nature, used across various fields for solving complex problems like machine learning and engineering design.
Post
Genetic Algorithm [GA]
GA is an optimization technique inspired by natural selection, using processes like selection, crossover, and mutation to evolve solutions over generations, with applications in areas like machine learning, design optimization, and AI behavior evolution.