SYSTEM STATUS ▸ ANDROID • PYTHON • AUTOMATION • ACTIVE
I build privacy-conscious, hardware-aware software across Android, Python automation, applied AI, and document workflows.
- 📱 Mobile & Edge Systems: Android applications using Kotlin, Jetpack Compose, LiteRT, Vulkan, and camera OCR.
- 🛡️ Applied AI & NLP: Retrieval workflows using BM25, vector embeddings, structured evaluation, and defensive input handling.
- ⚙️ Reliable Automation: Resilient media tools and reversible filesystem utilities designed for real-world failure conditions.
🌟 Start here:
Local-LLM-AI
My flagship Android project for exploring offline language-model execution, camera OCR, and performance on real mobile hardware.
| Project | Focus | Stack |
|---|---|---|
Local-LLM-AI |
Offline Android LLM and OCR workflows designed for reduced cloud dependence | Kotlin · Compose · LiteRT · Vulkan |
APEX-ATS |
Hybrid resume retrieval and screening safeguards | Python · BM25 · Embeddings · React |
PyStream-Downloader |
Resilient authorized stream capture | Python · Async I/O · HLS/DASH |
TubeNotes— Converts educational media into searchable, vector-crisp study-guide PDFs.PyFile-Organizer— Safe directory organization with non-destructive preview and undo manifests.Py-Image-Compressor— Batch image compression and format conversion with practical quality controls.Image-2-Pdf-Demo— Multi-image PDF generation for clean document workflows.Anime-Recommender— Gemini-powered recommendation workflow and content matching.
| Focus | Demonstrated Techniques & Capabilities |
|---|---|
| Mobile Performance | LiteRT runtime integration, Vulkan GPU acceleration, Android hardware profiling |
| AI Reliability | Prompt-injection testing, malformed-input sanitization, structured evaluation workflows |
| Automation Safety | Dry-run preview modes, operation logs, JSON rollback manifests |
| Document Workflows | Camera OCR pipelines, image preprocessing, quality-aware compression, and PDF generation |
| Resilient Systems | Non-blocking async I/O, exponential backoff retries, progress tracking, and failure recovery |
AI & Edge Runtime
Languages
Engineering & Frameworks
Systems & Applied Techniques
Vulkan · BM25 · Vector Retrieval · HLS/DASH · Async I/O · OCR Pipelines · PDF Generation
- Privacy by Architecture: Keep sensitive processing local where practical, minimize unnecessary data transfer, and make data handling explicit.
- Security & Robustness: Test AI-assisted workflows against prompt injection, malformed inputs, unsafe outputs, and predictable failure modes.
- Safe, Reversible Automation: Prefer previews, explicit apply steps, operation logs, and recovery paths for tools that modify user files.
- 🔭 Currently Exploring: Mobile performance profiling, model quantization, and robust evaluation of AI-assisted workflows.
- 🤝 Open To: Android, Python systems, applied NLP, developer tooling, and open-source collaboration.
- 🌐 Portfolio: badsiwal.my-style.in
- 💼 LinkedIn: Connect with me
Built with care • Focused on useful, reliable software