An AST-free, LLM-free heuristic knowledge graph engine for deep repository intelligence. Map, secure, and modernize enterprise codebases across 50+ languages at extreme velocity

"Adaptive Hybrid Quantization Framework for deploying 7B+ LLMs on low-VRAM devices (e.g., GTX 1050). Features surgical block alignment and Numba-accelerated inference.

Verified AI infrastructure for regulated deployment. UltraCompress (our wedge): near-lossless 5-bit compression with SHA-256-reproducible reconstruction - prove the model in production is the one you validated. 23 architectures (0.6B-405B), Hermes-3-405B @ 1.0066x. OpenAI-compatible API. pip install ultracompress

Project OpenSourceTerracore R&D open-source innovations

This repository contains our VXP2 Iteration 1.0B for IISc FSID/STEM Technical Assessment. Hybrid AI-Physics Predictive Maintenance for Aerospace. Integrating LSTM Neural Networks with Thermodynamic Guardrails to achieve >80% accuracy in engine RUL prediction. Built for Orion Spacetech.

Advanced Engineering Portfolio: Reference Implementations for Post-Quantum Security, Sovereign AI, and Kinetic Defense Systems. // 16 Scientific Modules.

omphalOS is an investigation platform for high-stakes analytical work — originally built for the U.S. Bureau of Industry and Security.

Deep-tech funding & valuation tracker for Indian startups — semiconductors, space, aerospace, eVTOL, biotech. Sector/quarter analytics, investor leaderboard.

Deep Tech Scout 是一款基於多 Agent 架構的深科技研發情資引擎。本工具先以以美國 NSF 科研專案為核心示範，整合全球專利分析框架、市場訊號（SEC 財報、ArXiv 論文）與即時產業新聞。透過自動化數據抓取與 LLM 深度分析，協助研究人員快速掌握研發趨勢、識別核心參與者，並自動生成繁體中文產業分析報告。

PyTorch-accelerated quantum simulation dashboard modeling multi-domain parameters (AgriTech VQE, Logistics QAOA, Finance ML-KEM) compiled for Tensor Core accelerators. Deployed via Streamlit for Quantumizers Meetup 1 at RTIH Visakhapatnam.

Score deep tech companies and their markets against a transparent, tunable rubric. LLM gathers facts; deterministic code scores. SEC EDGAR + Google Trends signals.

🚀 Run modern 7B LLMs on legacy 4GB GPUs without crashes, breaking the VRAM barrier for developers facing GPU limitations.