Our Work

Real Projects, Real Results

Computer vision, LLM agents and data annotation we have actually built and shipped, with live demo videos. Client names are withheld for confidentiality.

Real-Time Person Detection & Counting
Computer Vision

Real-Time Person Detection & Counting

Live RTMP camera streams processed with YOLOv8 to detect and count people frame by frame, served through a FastAPI + Docker inference service.

YOLOv8OpenCVFastAPIDocker
▶ Demo video
LLM & Agents

AI Chatbot Trained on Company Documents

RAG assistant that answers from PDFs and internal docs with source citations, and says "I don't know" instead of guessing.

RAGOpenAI / MistralVector DBPython
▶ Demo video
LLM & Agents

AI Agents & n8n Business Automation

Agentic workflows where an LLM decides the next step: triaging emails, qualifying leads and replacing recurring manual operations.

n8nLLM AgentsAPIsWebhooks
Retail, Warehouse & Product Datasets
Data Annotation

Retail, Warehouse & Product Datasets

18 production annotation projects and 45,000+ images in Roboflow: shelf products, CCTV, stacked sacks and crates, with boxes and polygons ready to train.

RoboflowCVATYOLO / COCOQA
Local AI Desktop Assistant
LLM & Agents

Local AI Desktop Assistant

Private, offline LLM assistant with system control and tool access through the Model Context Protocol. No data leaves the machine.

Llama 3.2llama.cppMCPPyQt5
Conversational Agent for Mental Health
LLM & Agents

Conversational Agent for Mental Health

Support agent combining rule-based logic, retrieval (LSTM/BiLSTM) and generative LLM responses with safety guardrails.

NLPLSTMGPTPython
AI-Powered Portfolio Website
Computer Vision

AI-Powered Portfolio Website

Web app running object detection, pose estimation and segmentation live in the browser, plus a voice AI assistant.

TensorFlow.jsGemini APIJavaScriptTailwind
See It In Action

Spectravise in Motion

Short demos of what we build: AI that answers from your documents and agents that automate real work.

Ready to build something your customers will love?

Tell us your problem and get a free proof-of-concept: what we would build, which stack, the risks and a realistic timeline. We usually reply within 24 hours.