Havuz

Personal academic and research portfolio focused on AI, low-resource Urdu NLP, RAG systems, machine learning, and computer science.

Native AI inference for PHP 8.3+ - run ONNX, GGUF (llama.cpp) and RubixML models directly in your PHP process via FFI. Chat, streaming, embe

The open, machine-readable standard for the Somali language — orthography, grammar, terminology, translation, and AI/benchmark resources, as

LLM-free, layout-aware PDF chunking for RAG pipelines. Preserves document structure via regex and font heuristics. Written in pure Rust.

Voynich Manuscript Decoded: Elu-Sinhala Phonetic Transcription & Vocabulary Toolkit 2026

Token Cost Parity: Multilingual LLM Efficiency Analysis 2026

RAGless is a semantic retrieval system that answers questions about your documentation, without using an LLM at runtime.

Spectral graph-augmented retrieval for RAG. Augments dense vector search with a weighted chunk graph and its Laplacian spectrum, graph expan

AI-detector auditing toolkit: measures how often a detector flags genuine human writing, how stable its verdict is across seeds, and whether

A pure-Python structured memory benchmark for multi-agent LLM systems — context graph vs vector RAG vs raw history dump, five scenarios, 18

Locally fine-tuned Russian RAG models: document splitter, query expansion, retrieval embedder. Teacher distillation → GGUF → llama.cpp on AM

Accepted at ISCSLP 2026 · Target-level benchmark for raw-input Chinese news TTS pronunciation

从 MiniMind 源码读起,再延伸到现代大模型技术体系的中文学习笔记。主线逐行精读预训练 / SFT / DPO / PPO / GRPO 与训练机制;附录 17 篇进阶卷覆盖量化、投机解码、RLHF 全景、模型代际史等 MiniMind 没涉及、但进阶绕不开的主题。

Train linear embedding adapters with triplet loss to align retrieval embeddings with your queries (RAG).

GateMem: a benchmark and evaluation toolkit for memory governance in multi-principal shared-memory LLM agents.

面向小团队的一站式 NLP 小模型平台:数据集版本管理 · 训练 · 评估 · 在线部署 · Badcase 闭环,内置大模型 Prompt 盲测评测(人工 + AI 双指标)。

GPT-2-style LLM built from scratch in C/CUDA with hand-written backprop, BPE tokenizer, FlashAttention, pretraining, and SFT.

Free, local-first academic integrity checker and MCP server: citation hallucination detection, AI-text and plagiarism signals, venue guideli

Somali NLP Corpus is an open, high-quality Somali text corpus designed to support research and development in Natural Language Processing (N

Salience-aware lexical trie for long-context compression.

Open benchmark for expert collapse and routing efficiency in sparse MoE LLMs

Hybrid RAG over Indonesian legal documents (UU, PP, Perpres, court decisions) with local LLM via Ollama.

211 free visual cards about large language models, from tokenization to deployment. Full collection: 376 cards.

Custom APRL machine-learning algorithm + a curated ML knowledge base of 923 papers, lectures, and explainers. NumPy classifier/regressor, te

A cautious, explainable AI text detector for local workflows and coding agents. Need production detection in your app? Try the Lynote AI Det

Caveman prompting, measured. A two-channel evaluation framework for prompt compression and LLM token efficiency: what input vs output compre

A structured knowledge base for large language model technologies, covering LLMs, Agents, RAG, model training, evaluation methodologies, and

Production-grade LLM Evaluation & Benchmarking Framework - GPT-4, Claude, Gemini, Mistral. Accuracy, latency, cost, hallucination, reasonin

AI tarafından üretilen metni LLM yeniden yazımları ve farklı makine çevirisi servisleriyle yeniden yapılandıran, yerelde çalıştırılabilir bir Python pipeline'ıdır.

Neden yükseliyor · 132 günde 2.979 yıldıza ulaşması belirgin ilgiye işaret ediyor; son 28 gündeki 21 commit ve 4 farklı yazar da projenin güncel olarak geliştirildiğini gösteriyor. Bu veriler ilgi ve aktivite kanıtıdır, çıktının kalitesi veya iddia edilen etkinliği için bağımsız kanıt değildir.

First open-source AI knowledge layer for Togo - 62K+ documents, RAG API, fine-tuned LLM. Built for developers, startups and institutions i

A modern, hands-on, self-paced course that takes you from your first line of Python to shipping a real AI-powered automation.

A lightweight LLM evaluation layer that turns subjective model outputs into deterministic decisions by scoring attribution, specificity, rel

LLM-powered conversation summarisation service for OCTO — turns group chats and threads into structured briefs with key decisions, open ques

This repository is a comprehensive, hands on roadmap to Python Mastery and AI Engineering. It is designed to take a learner from the very fo

Academic Data Scraper para extração massiva de metadados e abstracts (Scholar/Scielo/arXiv). Otimizado para alta performance e baixo consumo

📚 Read a top-tier AI paper in 5 minutes.

Better Estonian from your AI agent. Offline MCP server (no third-party calls) wrapping EstNLTK, EKI Reeglid + Riigi Teataja. 28 read-only to

Advanced RAG (Retrieval-Augmented Generation) system for private document intelligence. Uses OpenAI GPT-4o and Vector Databases to provide z

🏆 Fast, multilingual PII detection for privacy compliance (PDPA, PDPD, PDP Law, PIPL). 9 entity types, 13+ languages, CPU-only ONNX Runtime

日本語フリガナ (ルビ) を扱う Rust 製ライブラリ + ローカル HTTP サーバー。ルールはすべてデータ駆動 (TOML)。

A high-accuracy, multilingual rule-based Sentence Boundary Detector (SBD) with a drop-in adapter for pysbd, delivering faster and more accur

Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models

A post-retrieval temporal layer for RAG systems — validity filtering, time decay, document kind classification, and hybrid reranking in one

Humanize KR, Korean AI metinlerindeki çeviri kokan ve mekanik ifade kalıplarını tespit edip LLM yardımıyla daha doğal bir üslupla yeniden yazan bir CLI skill’idir.

Neden yükseliyor · 156 günde 5.711 yıldız alması belirgin ilgiye işaret ediyor; son 28 günde 57 commit ve 9 farklı yazar da projenin güncel katkı aldığını gösteriyor. Korean diline odaklanması ve birden fazla CLI ortamını desteklemesi ilgiyi artırmış olabilir; veriler tek başına yükselişin nedenini kanıtlamaz.

Turn documents into high-quality instruction datasets with grounding, quality filtering, deduplication, and provenance.

Discover the citing papers that matter most — ranked by impact, relevance, and influence

Open-source research intelligence for CS papers — track what matters, who drives it, what to read first, and where the research gaps are.

Online-adaptive multi-agent framework for enterprise API orchestration. Adaptive circuit breakers, SLA-driven routing, hybrid LLM+rule inten

🔍 Detect AI-generated text and fingerprint which LLM wrote it. Open-source GPTZero alternative. Zero dependencies, works offline.

A compounding LLM-wiki, best fit as a personal second brain that organizes and updates itself as it grows, zero upkeep.