AI that does the work.
Agents, custom models and WhatsApp automation. Built and secured by one team.
One team. Three disciplines.
AI is only as good as the software it runs on and the security around it. We own all three, so nothing falls between vendors.
Seven ways we put AI to work.
Most AI projects stop at a demo. Ours go into production, connected to your systems and secured before launch.
LLM fine-tuning
Models trained on your documents, tone and decisions.
LoRA adapters
Specialise open models fast, at a fraction of the cost.
AI agents
Assistants that take action across your tools.
WhatsApp AI agents
Bookings, orders and support on the app your customers use.
Agentic workflows
Multi-step automation with judgement built in.
RAG knowledge assistants
Answers grounded in your own documents, with sources.
Private LLM deployment
Models on your own servers. Data never leaves.
Live demo
Meet the agent that answers at 2 AM.
Pick a scenario and chat with a sample WhatsApp agent. Every reply does real work in your systems: bookings, records and reminders.
Interactive demo with sample data. Nothing is sent anywhere.
Workflows that run themselves, and know when to ask.
Agentic workflows chain AI decisions with the tools you already use, and stop for a person when the stakes are high.
Codexa builds what the AI runs on.
Our engineering studio designs and ships websites, web platforms, mobile apps and business systems: the foundations every AI project needs.
- Web platforms and e-commerce
- iOS, Android and cross-platform apps
- POS, CRM, ERP and admin systems
- Cloud, DevOps and long-term support
- Integrations that connect AI to your data
Nyoxa breaks it before attackers do.
Our offensive security practice tests web apps, APIs, cloud and AI systems, so what we build stays safe in the real world.
- Web application and API penetration testing
- Cloud and SaaS configuration review
- External attack surface assessment
- AI red-teaming: prompt injection and data-leak tests
- Fix-ready reports with free retests
From first call to production.
Listen
We learn the process, the data and the numbers you want to move.
Prove
A working pilot on your real examples, usually within weeks.
Build
Production integration, guardrails, monitoring and a Nyoxa security review.
Run
Support, retraining and monthly improvements as your business changes.
Questions we hear every week.
Is our data safe with AI?
Your data is used only for your project. We can keep it in your own cloud account or on your own servers, it is never used to train shared models, and every system gets a security review by Nyoxa before launch.
Do you support Sinhala and Tamil?
Yes. Our agents and assistants handle English, Sinhala and Tamil, including the mixed and transliterated messages people actually send.
Which AI models do you use?
We are model-agnostic: OpenAI, Anthropic Claude and Google Gemini, plus open models such as Llama, Mistral and Qwen. We choose per task on quality, cost and privacy, and you can switch later.
How long does a first project take?
A focused pilot, such as a WhatsApp agent or a knowledge assistant, typically goes live in three to six weeks. Larger fine-tuning or multi-system workflows take longer, and we scope them before you commit.
What does it cost?
It depends on scope, so we start with a fixed-price discovery and pilot. You see real results on your own data before any larger investment.
Put AI to work this quarter.
Tell us one process that eats your team's time. We will show you what an agent can do with it.