AI Development
Building on years of experience in software and embedded systems development, A-VEKT develops solutions powered by AI (artificial intelligence). We support the entire process, from clarifying requirements to selecting and implementing models and integrating them into your systems.
Focus Areas
- Natural language processing (chatbots, FAQ answering, document search)
- Retrieval-Augmented Generation (RAG) for internal documents and product information
- Application development using Large Language Models (LLM)
- Edge AI and machine learning leveraging our embedded and signal-processing expertise
- Integration of AI features and API connectivity into existing systems
Cost Reduction through AI Engine Optimization
Much of the cost of running AI comes from the per-token charges of the model (AI engine) you use. A-VEKT assesses the required accuracy and processing volume and selects the optimal model for each use case, delivering solutions that keep API costs down. By using higher-grade models only where high accuracy is needed and lightweight models for high-volume or routine processing, we can significantly reduce costs while maintaining quality.
The table below shows an example of how token pricing differs by model.
| Model | Typical Use | Input Price USD / 1M tokens |
Output Price USD / 1M tokens |
|---|---|---|---|
| GPT-5.6 | Complex reasoning and advanced text generation | $1.25 | $10.00 |
| GPT-5-mini | General chat and summarization | $0.25 | $2.00 |
| GPT-5-nano | High-volume processing, classification, simple replies | $0.05 | $0.40 |
* The prices above are illustrative examples. Please check each provider for the latest pricing.
In addition, because output tokens are priced higher than input tokens, reducing the amount of output leads to substantial cost savings. A-VEKT also proposes solutions that minimize cost by reducing unnecessary output through concise, structured prompt design and optimized processing flows.
Local LLM Support
For needs such as not being able to send data externally or wanting to avoid ongoing API charges, we also support deploying local LLMs that run on-premises or within your own environment. We can propose configurations that incur no usage-based API fees, or that keep confidential data entirely within your organization.
Development Results
In-house NuGet libraries (for .NET):
- Avekt.Ai.RagEngine — A RAG search engine combining keyword and embedding search with reranking. A provider-neutral core to which embedding providers (OpenAI, BGE-M3 on ONNX) and a BGE reranker can be added as plug-ins.
- Avekt.Ai.DocSearch — A document search library that extracts, chunks, and embeds PDF, Word, Excel, and text files in a folder and performs vector similarity search on SQL Server.
- Avekt.Ai.JsonSchemaGenerator — Generates JSON conforming to a given JSON Schema from natural-language instructions. Supports OpenAI, Azure OpenAI, and local LLMs (e.g., Ollama).
- Avekt.Ai.SmartFaqChat — A reusable ASP.NET Core FAQ chatbot module with RAG search, admin review, and inquiry storage. Used for the "SmartFAQ" AI chatbot on this site.
If you are considering adopting AI, please feel free to contact us through our inquiry form.