AI-powered platform for OSINT intelligence analysis. Features archive discovery with hypothesis-driven investigation, GLiNER entity extraction, Mapbox geospatial visualization, network analysis, and document processing. Built with FastAPI, Next.js, Weaviate, and DSPy.

Lightweight Python SDK for PII detection, redaction, and LLM guardrails, with fast regex defaults and optional NLP/OCR/Spark extras.

Production inference for encoder models - ColBERT, GLiNER, ColPali, embeddings etc. - as vLLM plugins for online and in-process deployment

Text preprocessing and PII anonymisation for NLP/ML. ONNX NER ensemble, language detection, stopword removal. Built for statistical ML and language models.

Collective AI memory for multi-agent teams. 10-stage GAMR search engine, knowledge graph (Apache AGE), document ingestion, 22+ MCP tools.

Comparison case study between Gliner V2 vs OpenAI Privacy filter model for PII Redaction

Easily configurable API & frontend providing simple access to dynamic NER models.

BrainUp is an AI-driven self-improvement app created for Hack the Hill 2024. It analyzes saved Instagram reels and posts to recommend productive activities and classes based on user interests. Built with ReactJS for the frontend and a Flask API for the backend, BrainUp leverages AI models like HDBSCAN, MPnet-base v2, and Mistral for data analysis,

Leverage ModernGLiNER's capabilities using LitServe.

🔎 NER (Named Entity Recognition) Application with Gliner 🛠️

This project focuses on developing and fine-tuning models for medical term extraction and general named entity recognition.

Identifies legal, regulatory, and procedural mechanisms used to consolidate power — and generates ranked, plain-English countermeasures. BYOK (bring your own Anthropic key). Powered by GLiNER2 + Claude.

Clause-level extraction from legal contracts using NER and structured prompting.

Standalone RAG backend: hybrid search (pgvector + ParadeDB BM25 + LightRAG graph) with 4-mode rerank (CPU TEI + AMD gfx1151 GPU sidecars: BGE, Qwen3-Reranker 4B/8B) and GLiNER NER fast-mode. FastAPI on :5050. Bind-mount persistence.

Comparative study of parameter-efficient fine-tuning (PEFT) strategies for biomedical NER on top of GLiNER — including soft prompt tuning, embedding injection, and a custom in-place embedding extension that matches full fine-tuning performance at 13% of trainable parameters.

High-performance Hybrid RAG pipeline using LlamaIndex, MinerU, GLiNER, and LangExtract. Features advanced metadata enrichment, BM25-Vector fusion, and LaTeX-enabled synthesis for complex document analysis.

Fine-tuning workflows for GLiNER 2 on domain-specific entity recognition.

MLX support for gliner2

🛡️ Text Anonymizer using NuNER Zero-shot An academic project that auto-detects & masks sensitive entities (names, orgs, locations) in .txt/.docx files using Zero-shot NER. Features CLI & REST API. Replaces data with realistic fakes (Faker) or [REDACTED]. Built with Python, Flask, GLiNER.