Privacy isn’t a feature. It’s a foundation.
I design and build reliable, privacy-first systems with a strong focus on clarity, automation, and long-term maintainability.
My work sits at the intersection of software engineering, data integrity, and domain-specific AI, with a strong emphasis on user trust, platform safety, and long-term system health.
I believe good systems should be:
- Transparent by design
- Predictable in behavior
- Respectful of user data
- Quietly effective rather than attention-seeking
NoBullFit is a privacy-first health and fitness tracking platform that I actively design, build, and operate.
It allows users to track nutrition, recipes, activities, weight, and wellness data while maintaining full ownership of their information.
Core features remain free, and data is never sold or exploited.
- Live platform: https://nobull.fit
- Web application repository:
https://github.com/pathvoid/nobullfit
Key characteristics
- Modern web architecture (React, TypeScript, Express, PostgreSQL)
- Server-side rendering for performance and reliability
- Full data export and deletion capabilities
- Clear separation between product logic and background processing
- Strong privacy and data minimization principles
To support platform quality and safety, NoBullFit includes a dedicated AI background worker responsible for automated verification and moderation.
- Worker repository:
https://github.com/pathvoid/nobullfit-worker
What it handles
- AI-based content verification
- Text and image moderation
- Illegal content detection and automatic containment
- Controlled scheduling to avoid over-processing
- Human escalation only when necessary
Design goals
- Safety without surveillance
- Automation without overreach
- Clear auditability of decisions
- Independent operation from the main web application
NoBullFit AI is a custom-trained language model specialized in fitness, health, nutrition, and mathematical conversions.
Unlike general-purpose models, it is trained from scratch on domain-specific data, with fully automated data collection and resumable training.
- AI repository:
https://github.com/pathvoid/nobullfit-ai
Key characteristics
- GPT-2–based transformer architecture
- Automated web data collection and synthetic data generation
- Domain-specific training for fitness, health, and nutrition use cases
- Built-in checkpointing with pause and resume support
- Local training with no dependency on external cloud services
Purpose
- Provide accurate, context-aware responses within the NoBullFit ecosystem
- Reduce reliance on generic AI models
- Maintain privacy and control over the full AI training lifecycle
Privacy-first, always
User data belongs to the user. Privacy is treated as infrastructure, not a premium feature.
Automation with accountability
AI is used to reduce manual workload, not to obscure decisions.
Systems over shortcuts
I prioritize architectures that scale in clarity before they scale in size.
Professional seriousness
I build tools meant to be maintained, audited, and trusted over time.
- Backend & APIs (TypeScript, Python, PostgreSQL)
- Background workers & automation pipelines
- Domain-specific AI and custom model training
- Content moderation and platform safety systems
- Data integrity, lifecycle management, and auditability
- Clean system boundaries and responsibility separation
- Expanding domain-specific AI capabilities within NoBullFit
- Strengthening automated verification and safety tooling
- Improving long-term maintainability and observability
- Building sustainable, self-hostable, privacy-respecting systems
- Applying these principles across health, productivity, and internal tooling
If you’re reviewing this profile as a recruiter, collaborator, or engineer:
what you see here reflects how I work in real systems - quietly, deliberately, and with intent.