Konko Maji

Exploring computational linguistics to preserve and advance the Bengali language in the AI era.

Konko Maji is an undergrad from West Bengal. This site traces back to one question: how does a language as old as Bengali hold up inside AI models that were never built with it in mind.

Background

No computer science degree, currently reading Accountancy Honours. Everything technical here is self taught, built outside class hours. Before that, close to three years of hands on Google Ads and Meta Ads media buying and digital advertising strategy. Also into economics, statistics, and political science, mostly applied to West Bengal and India.

Bengali NLP research

  • bornomala Bengali-first, dialect-aware LLM research and open tooling. Includes a Bengali tokenizer that beats Sarvam, AI4Bharat, and GPT-4o, fertility 1.39, near-zero broken conjuncts.
  • MotherTongueIndex Multilingual tokenizer efficiency analyzer, measures how many more tokens your language costs than English.

Civic & public data

  • betar Offline mesh messaging and safety app, chat, SOS, disaster bulletins, offline maps. No SIM, no internet, no server. Built on Project Mesh.
  • indianjobsmap D3 treemap of India's 61.6 crore workforce by occupation and AI exposure, built on PLFS 2025 microdata.
  • westbengal Interactive explorer of the West Bengal state budget 2026-27, built from the official Finance Department publications.
  • wb-govt-jobs-alert Background scraper and alert system for West Bengal government job notifications across 68 official portals.
  • fifa-world-cup-2026-final-prediction Statistical prediction model for the 2026 World Cup final, nine models, 400,000 Monte Carlo simulations on official FIFA tracking data.

Tools & infrastructure

  • wordpress-mcp Universal MCP server for WordPress, connects any site to Claude for hands-on SEO/AEO/GEO work, engine-agnostic for Yoast or RankMath.
  • webtoolshub 12 free AI-native domain and website tools, WHOIS, DNS, SSL, SPF/DMARC, security headers, usable via MCP or REST. Stdlib Python, no signup.
  • wp-certificates Open-source, white-label certificate engine for WordPress, visual builder, verifiable PDFs, Tutor LMS and WooCommerce hooks.
  • gmb-autopilot Full-stack Google Business Profile automation engine, AI reviews, content, and SEO optimizer for agencies.
  • wp-seo-hub Open-source WordPress SEO dashboard, multi-site meta optimization with Claude, Google Indexing API, keyword tracking. Local only, data never leaves your machine.
  • geotagcamera Free, ad-free Android app that stamps photos with GPS location, address, and timestamp. Built for field surveyors and NGOs.