This category explores various programming concepts and from web development to mobile development. This document my programming journey from novice to principal computer scientist.
Google Research’s TurboQuant introduces an innovative solution to the memory challenges of large language models with its PolarQuant technique. By significantly compressing the key-value cache without sacrificing quality, it enhances efficiency and speeds up responses. This advancement opens doors for high-quality long-context AI applications on accessible hardware. Exciting times ahead!
We took ColPali (vision-language embeddings for documents) and Light-ColPali (token merging via hierarchical clustering) and built the production infrastructure around them. The system uses PostgreSQL + pgvector as a unified store, a lease-based job queue for resilient ingestion, and a two-stage retrieval pipeline that retrieves at patch granularity but ranks at page level.
The key insight: text extraction is lossy. For documents with complex layouts, charts, and tables, embedding the rendered page as an image solves problems that text-based RAG can’t touch.
Transformer models revolutionised AI by replacing recurrence with attention, allowing machines to understand context, relationships, and meaning across long sequences of data. This post explores how GPT models harness the Transformer architecture to generate coherent, intelligent text and why this innovation reshaped the entire AI landscape.
Semantic chunking is a smarter, meaning-based approach to breaking text into pieces for Retrieval-Augmented Generation (RAG) systems. Instead of slicing by token count, it groups sentences that belong together, keeping the logic and flow intact. This helps retrieval systems pull coherent context rather than fragmented thoughts—a crucial advantage when dealing with complex documents like contracts or financial statements. While current research shows mixed evidence on its effectiveness, the theoretical promise is clear: semantic chunking preserves context, coherence, and clarity, making RAG systems feel less mechanical and more genuinely understanding.
This post explores how DeepSeek-OCR’s visual context compression could expand LLM memory and inspire new multimodal thinking, drawing on chain-of-thought training ideas.
Alphora brings institutional portfolio analytics to long-term investors: MPT optimizer, risk-adjusted metrics, dividend intelligence, and portfolio-aware news. Join the waitlist.
I’m currently juggling a full-time job while building a platform for mid-tier influencers; inspired by my brother’s struggles...
Introduction What happens when you train a character-level GPT model on the entire works of William Shakespeare? You...
If you have ever owned a self help book, scrolled too far on instagram or listend to any...
After considering the look and feel of the application, I have decided to update the UI/UX of my...
Leading up to Nvidia’s Q4 2024 earnings call on February 22nd, the stock witnessed a significant upsurge. Following...
Currently, I’m immersed in creating a goal-tracking application designed specifically for St Thomas the Apostle College. This project...
In the bustling world of technology, where every new device promises to be the next big revolution, the...
Github Repository 2023 – Present | Full-stack development | Technologies: Python, Flask, reactjs, Openai API, FFmpeg, pyDub, mpviepy,...
AI / Programming / Tech News
DeepSeek-OCR and the Future of AI Memory and Visual Thought
October 21, 2025
