<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Tanvir Ahmed Sijan</title><link>https://sijantanvir.github.io/</link><atom:link href="https://sijantanvir.github.io/index.xml" rel="self" type="application/rss+xml"/><description>Tanvir Ahmed Sijan</description><generator>Wowchemy (https://wowchemy.com)</generator><language>en-us</language><copyright>Copyright © 2026</copyright><lastBuildDate>Sun, 25 Oct 2026 00:08:00 +0000</lastBuildDate><image><url>https://sijantanvir.github.io/media/icon_hude4d2d53747fe99a47569dfb0040aebb_6903_512x512_fill_lanczos_center_3.png</url><title>Tanvir Ahmed Sijan</title><link>https://sijantanvir.github.io/</link></image><item><title>Are Arithmetic Heuristic Neurons Form-Invariant? A Mechanistic Analysis of Symbols, Text, and Code in LLMs</title><link>https://sijantanvir.github.io/publication/2026-arithmaticformat/</link><pubDate>Sun, 25 Oct 2026 00:08:00 +0000</pubDate><guid>https://sijantanvir.github.io/publication/2026-arithmaticformat/</guid><description/></item><item><title>Beyond Clean Text: Evaluating Encoder and Decoder Robustness for Bangla Event Detection in Noisy Text</title><link>https://sijantanvir.github.io/publication/2026-banglanoisyevent/</link><pubDate>Sun, 25 Oct 2026 00:01:00 +0000</pubDate><guid>https://sijantanvir.github.io/publication/2026-banglanoisyevent/</guid><description/></item><item><title>Implementing Transformers from Scratch</title><link>https://sijantanvir.github.io/project-tutorials/transformers-from-scratch/</link><pubDate>Sun, 02 Aug 2026 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/project-tutorials/transformers-from-scratch/</guid><description>&lt;hr>
&lt;p>This is a work-in-progress project where I build a complete Transformer pipeline from the ground up. The goal is to implement one component at a time, in order: tokenizer → embeddings → positional encoding → attention → feed-forward → layer norm → encoder block → full model → loss → training → evaluation.&lt;/p></description></item><item><title>Implementing DPO from Scratch</title><link>https://sijantanvir.github.io/project-tutorials/dpo-from-scratch/</link><pubDate>Mon, 22 Jun 2026 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/project-tutorials/dpo-from-scratch/</guid><description>&lt;hr>
&lt;p>An implementation of Direct Preference Optimization (DPO) for language model alignment using PyTorch. without Hugging Face TRL.&lt;/p></description></item><item><title>BanglaSocialBench: A Benchmark for Evaluating Sociopragmatic and Cultural Alignment of LLMs in Bangladeshi Social Interaction</title><link>https://sijantanvir.github.io/publication/2026-banglasocialbench/</link><pubDate>Mon, 16 Mar 2026 21:58:32 +0000</pubDate><guid>https://sijantanvir.github.io/publication/2026-banglasocialbench/</guid><description/></item><item><title>Assessing the Visual and Built Environment Risk on Pedestrian-Vehicle Crash Severity in Absence of Detailed Roadway Inventory: A Deep Learning Based Framework Harnessing Public Geospatial Data</title><link>https://sijantanvir.github.io/publication/2025-crashseverityml/</link><pubDate>Sun, 29 Jun 2025 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/publication/2025-crashseverityml/</guid><description/></item><item><title>Crash Severity Analysis with Public Geospatial Data</title><link>https://sijantanvir.github.io/project-tutorials/crash-built-visual-ml/</link><pubDate>Sat, 22 Feb 2025 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/project-tutorials/crash-built-visual-ml/</guid><description>&lt;p>In this project, I demonstrate how publicly available geospatial data and Google Streetview Images can be utilized to determine the correleation between fatal crashes and driver’s visual and built environment for data-scarce regions.&lt;/p>
&lt;p>I designed a pipeline that collects Point-of-Interest data from Overture Maps, collects information about road and traffic from Google Traffic Color Codes, weather data from Open-Meteo API and then perform Semantic Segmentation to identify Streetscape Complexity, Street Openness, Building View Index, Greenview Index. Then train four ensemble models (XGBoost, AdaBoost, Random Forest, CatBoost) to predict crash severity. Finally perform SHAP interpretability to map how individual road segments contrbute to fatal road crashes.&lt;/p>
&lt;p>&lt;img src="./assets/Figure_4.png" alt="Pipeline">&lt;/p></description></item><item><title/><link>https://sijantanvir.github.io/admin/config.yml</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/admin/config.yml</guid><description/></item><item><title/><link>https://sijantanvir.github.io/contact/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/contact/</guid><description/></item><item><title/><link>https://sijantanvir.github.io/projects/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/projects/</guid><description/></item><item><title/><link>https://sijantanvir.github.io/tutorials/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/tutorials/</guid><description/></item><item><title>Count</title><link>https://sijantanvir.github.io/count/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://sijantanvir.github.io/count/</guid><description/></item></channel></rss>