Fahd Mirza

Fahd Mirza is an AI researcher and content creator specializing in efficient local AI, model-compression, inference-optimization, and agentic systems. He is known for his work on ternary-quantization, speculative-decoding, and multimodal evaluation.

Key Projects & Research

PG-Jev

Fahd developed pg-jev, a PostgreSQL extension that enables natural-language-querying directly within SQL. This tool allows for filtering, ranking, and classifying database rows using plain English conditions, serving as a practical application for decision models and structured decision making.

• Focus: Bridging the gap between natural-language-processing and relational databases.
• Implementation: pg-jev acts as a bridge for ai-driven-decision-modeling in SQL environments.
• Related Work: See PG-Jev: Natural Language Querying and AI-Driven Decision Modeling for PostgreSQL for detailed technical notes.

Technical Interests & Tools

• Local AI Ecosystem: llamacpp, ollama, lm-studio, gguf formats.
• Model Families: qwen, deepseek, mistral-ai, claude-sonnet-55, nous-research (Hermes).
• Optimization Techniques: model-compression, ternary-quantization, sherry-quantization, angelslim, hy4, op55.
• Agentic Frameworks: agent-harness, hermes-agent, openjev, kev, jev, skill-spector.
• Hardware & Efficiency: energy-efficient hardware, neuromorphic computing, p-bits, gpu-poor workflows.
• Data & Privacy: data privacy, ccpa, gdpr, unbroker, data brokers.

References

• PG-Jev: Natural Language Querying and AI-Driven Decision Modeling for PostgreSQL

Source Notes

• 2026-10-10: PG-Jev: Natural Language Querying and AI-Driven Decision Modeling for PostgreSQL · ▶ source
• 2026-09-29: Claude Sonnet 5.5: Performance Benchmarks, Cost Efficiency, and Agentic Coding · ▶ source
• 2026-09-27: AI Decision Model Evaluation: Customer Service Urgency & Frustration Assessment · ▶ source
• 2026-09-24: AI Model Comparison: Espresso Machine Simulation & Latte Art Physics · ▶ source
• 2026-09-18: Tencent AI Model Shrink with Sherry Quantization: AngelSlim Breakthrough Report · ▶ source
• 2026-09-17: Jev: RLCD’s Shift from Human-Preferred Text to Calibrated Decisions · ▶ source
• 2026-09-09: NVIDIA SkillSpector: Scanning AI Agent Skills for Malware and Vulnerabilities · ▶ source
• 2026-09-08: Qwen3.8-27B Quantization: GSQ+RCO for Local, Accurate LLM Deployment · ▶ source
• 2026-09-07: OmegaClaw: A Neural-Symbolic AI Agent for Persistent, Explainable Reasoning · ▶ source
• 2026-09-06: Extropic Z1T: Probabilistic P-bits for 100x More Energy-Efficient AI Hardware · ▶ source
• 2026-09-04: Muse Spark 1.3: Meta’s Open-Weight Multimodal AI for Long-Horizon Agentic Work · ▶ source
• 2026-08-22: DeepSeek V4-Flash Vision Model Evaluation: Capabilities, Limitations, and Reasoning · ▶ source
• 2026-08-19: loop Feature: Automated Recurring Task Demo · ▶ source
• 2026-08-17: DeepSeek Harness: Local LLM Agent with Environment Interaction & Plugins · ▶ source
• 2026-08-15: Qwen3.8-27B Local Deployment and Performance Analysis Report · ▶ source
• 2026-08-15: Qwen 3.8-27B GGUF Local Deployment via llama.cpp, Ollama, LM Studio · ▶ source
• 2026-08-11: Muse Glimmer 30B: Meta’s Open Agentic Multimodal Model for Local AI · ▶ source
• 2026-08-10: Neutrino-8B: Ternary Quantization and Speculative Decoding for Efficient Local AI · ▶ source
• 2026-08-02: Local AI-Powered Note Management: Hermes Agent, Obsidian, Ollama Integration · ▶ source
• 2026-07-02: LongCat 2.0: China’s Nvidia-Free 1.6T AI Model Achieves Top Performance · ▶ source
• 2026-06-27: OpenJarvis: Stanford’s Local AI Framework for On-Device Privacy and Efficiency · ▶ source
• 2026-06-20: Fine-Tuned Qwen3.6-27B Pi-Reasoning GGUF for Local Agentic Code Debugging · ▶ source
• 2026-06-14: Key AI Developments: Claude Fable 5 Export Controls, DiffusionGemma, Coding Agents · ▶ source
• 2026-06-10: Google QAT vs. Unsloth QAT: Gemma 4 12B Performance Comparison · ▶ source
• 2026-06-10: Gemma 4 Chat Template Fix: Preserving Reasoning for Enhanced Agentic Performance · ▶ source
• 2026-06-09: Gemma 4 Was Broken for Agents - Google Just Fixed It · ▶ source
• 2026-06-06: Miso TTS 8B Emotive Text-to-Speech Model: Installation and Performance Review · ▶ source
• 2026-06-05: NVIDIA Nemotron 3 Ultra: Open LLM Agent Optimizes Fast API Performance · ▶ source
• 2026-06-03: Adaptive PFlash and Hermes Agent: Self-Tuning LLM Prefill for Long Contexts · ▶ source
• 2026-06-02: NVIDIA Cosmos 3: Omnimodal World Model for Physical AI Robotics · ▶ source
• 2026-06-01: MiniMax M3: Open-Weight LLM’s Frontier Coding, Native Multimodality, and Sparse Attention · ▶ source
• 2026-05-30: Weekly AI Developments: Opus 4.8, Step Audio 3, Bonsai Image (May 2026) · ▶ source
• 2026-05-28: DwarfStar: Native DeepSeek V4 Flash Local Inference with Persistent KV Cache · ▶ source
• 2026-05-28: Bonsai Image: Local 1-bit Image Generation Through Extreme Quantization · ▶ source
• 2026-05-26: MiniCPM-1B: Efficient 1B-Parameter LLM for On-Device Hybrid Reasoning · ▶ source
• 2026-05-22: llama.cpp Router Mode: Native Hot-Swappable Local LLM Switching · ▶ source
• 2026-05-22: EdgeQuake: Local Rust Graph-RAG with Ollama for Improved Knowledge Retrieval · ▶ source
• 2026-05-20: MTP + Ngram Stacked Speculative Decoding in Llama.cpp for LLM Inference · ▶ source
• 2026-05-18: GPU Deployment via llama.cpp Quantization · ▶ source
• 2026-05-13: TurboQuant & DFlash: Accelerating Local LLM Inference with Enhanced Context · ▶ source
• 2026-05-11: NVIDIA Nemotron Elastic: Bundling Three LLMs for Flexible Deployment · ▶ source
• 2026-05-10: ERNIE 5.1: Baidu’s AI Model - High Performance, Cost-Efficient, Multimodal Capabilities · ▶ source
• 2026-05-09: Local Deep Research: Local AI Assistant for Comprehensive Private Information Synthesis · ▶ source
• 2026-05-06: Google Gemma 4 MTP Drafters: Accelerating Inference Speed with Speculative Decoding · ▶ source
• 2026-05-03: Luce PFlash: 10x Faster AI Model Prompt Prefill on Local GPUs · ▶ source
• 2026-04-19: Qwen 36-35B Full Precision vs Ollama Quantized Performance-Memory Trade-off · ▶ source
• 2026-04-12: MiniMax M27 Open Source LLM Technical Overview and Deployment Summary · ▶ source
• 2026-04-10: Gemma 4-E2B LLM Fine-Tuning Custom Dataset Unsloth Local Tutorial · ▶ source
• 2026-04-08: Gemma 4-E2B LLM Fine-Tuning: Custom Dataset & Unsloth Local Tutorial · ▶ source
• 2026-04-07: Gemma 4-E2B LLM Fine-Tuning: Custom Dataset & Unsloth Local Tutorial · ▶ source