Skip to main content
  1. Blog
LinkedIn

Blog | Rantideb Howlader

Ranti

Rantideb Howlader

@ranti

Connect
Search PostsReading ListTimelineBlog Stats

Posts

L

Logging Off For A While

A short, quiet note about stepping away for a while and holding onto small things.

May 5, 2026 (1mo ago)8 min readPersonal
122
I

Implementing Grace: A PyTorch Case Study in Dual-Stream Dysfluency Models

The practical companion to The Topography of Hesitation. A technical case study detailing the implementation of the dual-stream GraceJoin architecture, Neural ODE effort streams, and integrating hesitation topology into LLM agents via MCP.

June 22, 2026 (3d ago)39 min readPyTorch
71
Forgetting Is Not Deletion: The Verification Gap in Machine Unlearning
Research

Forgetting Is Not Deletion: The Verification Gap in Machine Unlearning

An empirical and conceptual critique of the 'verification gap' in machine unlearning, demonstrating that passing a membership audit does not imply actual forgetting of generalizable relationships.

June 20, 2026 (5d ago)15 min readMachine Unlearning
60
T

The Topography of Hesitation: Non-Markovian Ruptures and the Mathematical Failure of Autoregressive Models on Dysfluent Speech

A formal argument that autoregressive speech models do not merely struggle with stuttered speech but are mathematically mandated to erase it, and a proposed continuous-time latent architecture that encodes hesitation as effort rather than collapsing it into a void.

June 18, 2026 (1w ago)42 min readMachine Learning
20
B

Beyond RAG: Using Multi-Agent Systems for Deep Cultural and Literary Analysis

An algorithmic breakdown of why dense vector retrieval fails on narrative structures, and how to architect stateful, graph-based Multi-Agent Systems for rigorous computational literary analysis, featuring a practical case study on Rabindranath Tagore.

June 15, 2026 (1w ago)33 min readMachine Learning
10
R

Running Local LLM Agents in Kubernetes: A Practitioner's Guide to vLLM on EKS

A hands-on guide to deploying high-throughput, self-hosted LLMs using vLLM on Amazon EKS, covering GPU node groups, autoscaling, and performance tuning.

June 12, 2026 (1w ago)28 min readAWS
10
A

Agent Looping and Systems Engineering: Building Reliable AI

Prompt engineering is a temporary fix for AI slop. True reliability comes from systems engineering. This post explores agent looping, Eval gates, and the GBRAIN architecture.

June 9, 2026 (2w ago)20 min read
30
B

Building a Multi-Writer Serverless SQLite Engine on Amazon S3

How to run SQLite in WAL mode on Amazon S3 for serverless applications.

June 5, 2026 (2w ago)91 min readEngineering
40
Letter to EveryOne

Letter to EveryOne

Dying Is an art, like everything else. I do it exceptionally well. I do it so it feels like hell. I do it so it feels real. I guess you could say I have a call.

June 3, 2026 (3w ago)17 min readLiterature
41
S

System Design Basics Databases Message Queues and Authentication

A complete deep dive into databases message queues authentication and proxies. All the core system design concepts explained simply with real examples and diagrams.

June 1, 2026 (3w ago)14 min readEngineering
30
S

System Design Basics Caching How to Make Your App Fast

Caching stores frequently used data in fast storage so your app does not hit the database every single time. Here is how it works from scratch with real examples.

May 28, 2026 (4w ago)19 min readEngineering
30
S

System Design Basics Scalability and Load Balancing

Breaking down system design concepts like scaling load balancers and redundancy into plain English. No jargon just common sense.

May 27, 2026 (4w ago)14 min readEngineering
00
•••