AIGNITE OS
aignite-os v1.0 · igniting…
ignite( ideas ) → working AI
AIGNITE is an AI development company based in Hải Phòng, Vietnam. We design intelligent products and apply AI to real problems — from drug discovery and climate tech to agriculture, education, and public services.
Everything in this OS is evidence, not promise: products we run, systems we've built, and domains we've shipped into. Concrete capabilities over hype.
capabilities · evidenced by shipped projects
Multi-provider integration, RAG, schema-constrained output, real-time audio, prompt-cache cost engineering, prompt-injection defense-in-depth.
PyTorch multi-task networks, gradient boosting, time-series forecasting, SHAP interpretability, small-data rigor.
YOLOv8 detection & segmentation, edge + cloud inference, classical CV, real-time DSP at interactive frame rates.
Molecular docking, ADMET, de novo design, adsorption-cycle simulation, techno-economic optimization, physics + ML hybrids.
Tauri desktop with embedded Python, React web, React Native mobile, offline-first PWAs, i18n.
MCU firmware with fail-safes, RTK GNSS, checksummed UART, on-device inference.
Cloud Run, Cloudflare Workers with geo-gating and rate limiting, size-optimized Docker, serverless crawlers.
selected work · sector level, client details confidential
biomedical · pharma r&d
Multi-task deep-learning resistance prediction with molecular docking, ADMET screening, and fragment-based de novo design — packaged as a cross-platform desktop app with an embedded scientific Python runtime.
climate tech · ccu
Hybrid physics + ML: models over experimental catalysis data, an adsorption-cycle simulator, and a techno-economic optimizer in a bilingual cloud web app with emissions forecasting.
precision agriculture
A field robot with RTK positioning streaming crop imagery to cloud YOLOv8 inference, a real-time dashboard, an advisory engine, and mobile push alerts — from firmware to cloud.
education · music
Schema-constrained LLM lesson generation with real-time browser audio: live pitch detection, streaming to multimodal models, and a multi-instrument simulator.
civic services · client work
Deterministic offline instruction layer plus a geo-gated, rate-limited LLM chat backend on edge serverless — provider-swappable models, lexical RAG cutting prompt tokens 4–10×, defense-in-depth output filtering.
ignite( your idea )
Tell us what you're trying to build — we'll tell you honestly whether AI is the right tool.
AIGNITE CO., LTD · Business reg. 0202277533 · Hải Phòng, Vietnam