Jsaha.in BLOGS
Portfolio
ENGINEERING, AI & SYSTEM ARCHITECTURE

Blogs & Articles

Practical writing, research perspectives, and technical commentary connecting software engineering, scalable AI systems, cloud security, and distributed architectures.

Explore Blogs XML Feed
ENGINEERING NOTES · LIVE FEED AI / ML 8 MIN NEW CYBERSECURITY 14 MIN ARCHITECTURE 10 MIN TOPIC INDEX AI & ML CLOUD SECURITY SYSTEMS DESIGN FULL-STACK XML.FEED $ tail -f /feed/blogs.xml + NEW POST
Spotlight

Featured Blogs

PHP2SHELL · FROM MEMORY BUG TO SHELL PHP WORKER MEMORY in use freed (stale) attacker data 1 · MEMORY RUNS OUT 2 · CLEANUP SKIPPED 3 · STALE DATA REUSED 4 · ATTACKER GETS SHELL $ id uid=33(www-data) ← no password needed CVSS 9.2 · CRITICAL NO LOGIN NEEDED ~70% OF THE WEB
Cybersecurity
Nov 01, 2026
J
Jyotirmoy Saha 9 min

PHP2Shell: The Bug That Put 70% of the Web at Risk

PHP2Shell (CVE-2026-109618 and CVE-2026-109619) lets attackers run code on PHP web servers without logging in. Plain-English timeline, animated diagrams, real exposure data, and a step-by-step patch plan.

PIPELINE ARCHITECTURE // RETRIEVAL-AUGMENTED GENERATION P@10: 96.8% | LATENCY: -68% 01. VECTOR.DB // HNSW INGEST • CHUNKING C_01 C_02 C_03 1536-D LATENT SPACE q_vec SQ8 -74% RAM FOOTPRINT DENSE BM25 02. HYBRID RRF FUSION SPARSE BM25 SCORES RRF PRISM & RERANK RRF = \Sigma 1/(60 + r_m) CROSS-ENCODER TOP-K=20 TOP 20 03. CONTEXT & LLM SMALL-TO-BIG EXPAND PARENT WINDOW HIT CHUNK STREAMING TOKENS token_0 token_1 token_2 GEN... P99 -68% LATENCY 1. EMBED 2. HNSW+BM25 3. RRF FUSION 4. LLM SYNTHESIS
AI & Machine Learning
Oct 30, 2026
J
Jyotirmoy Saha 8 min

Building Scalable Retrieval-Augmented Generation (RAG) Architectures

Architectural principles for combining vector databases, embeddings, and real-time enterprise knowledge retrieval.

ZERO-TRUST GATES · EVERY HOP VERIFIED CLOUD.A CLOUD.B ON-PREM IDENTITYPOLICYmTLS API WORKER DATA DENY > id=orders-api act=reserve ALLOW > id=batch-job act=export DENY > id=billing-wkr act=read ALLOW > id=orders-api act=reserve ALLOW > id=batch-job act=export DENY > id=billing-wkr act=read ALLOW AUDIT EVIDENCE
Cybersecurity
Oct 26, 2026
J
Jyotirmoy Saha 16 min

Implementing Zero-Trust Architecture in Multi-Cloud Infrastructure

Practical defense-in-depth engineering strategies for Kubernetes, Linux environments, and API gateways.

Complete Archive

All Blogs

PIPELINE ARCHITECTURE // RETRIEVAL-AUGMENTED GENERATION P@10: 96.8% | LATENCY: -68% 01. VECTOR.DB // HNSW INGEST • CHUNKING C_01 C_02 C_03 1536-D LATENT SPACE q_vec SQ8 -74% RAM FOOTPRINT DENSE BM25 02. HYBRID RRF FUSION SPARSE BM25 SCORES RRF PRISM & RERANK RRF = \Sigma 1/(60 + r_m) CROSS-ENCODER TOP-K=20 TOP 20 03. CONTEXT & LLM SMALL-TO-BIG EXPAND PARENT WINDOW HIT CHUNK STREAMING TOKENS token_0 token_1 token_2 GEN... P99 -68% LATENCY 1. EMBED 2. HNSW+BM25 3. RRF FUSION 4. LLM SYNTHESIS
AI & Machine Learning
Oct 30, 2026
J
Jyotirmoy Saha 8 min

Building Scalable Retrieval-Augmented Generation (RAG) Architectures

Architectural principles for combining vector databases, embeddings, and real-time enterprise knowledge retrieval.

Stay Updated

Subscribe for

Newsletter

Never miss a new article. Get engineering perspectives, AI pipelines, and code breakdowns straight to your inbox.

Zero spam. Strictly engineering & technology updates.