We build real software — not slides. Below are some of our own projects and work that show how Dharma solves real problems.
FraudTalon
Problem: digital scams, phishing and fraudulent messages arrive faster than people and businesses can analyze them.
Solution: a proprietary platform for analyzing suspicious URLs, emails, messages, images and files with AI, identifying fraud signals before the loss happens.
Stack: Python · spaCy · LLM
Log Analyzer
Problem: application and infrastructure logs grow fast and hide important signals of errors, abuse and attacks.
Solution: a tool to collect, normalize and analyze logs, surfacing anomalies and patterns relevant to security and operations.
Stack: Java · log processing · pattern detection
SovereignRAG
Problem: auditing code for security flaws usually means sending sensitive code to cloud-based AI services.
Solution: an open-source sovereign RAG engine for security code auditing, 100% offline, running on personal hardware (Apple M1 fully supported). Uses LLMs like phi3, mistral and codellama via Ollama, a vector DB with ChromaDB and preprocessing with spaCy for efficient semantic chunking.
Stack: Ollama (phi3 · mistral · codellama) · ChromaDB · spaCy
License: BSD 3-Clause
Custom work
We also build systems, APIs, integrations and automations for clients — in sectors such as banking, fintech and e-commerce.
For confidentiality reasons, we don’t expose client details. The focus stays on what matters: the problem solved and the technology applied.
- High-load backends and APIs
- Financial systems and critical flows
- Automation of manual processes
- System-to-system integrations
- Log analysis and application security
Have a similar problem? Discuss a project