I'm a Software Developer with 2+ years of experience building production SaaS applications, full-stack systems, and AI-powered business solutions.
Currently, I work at JWeis Marketing, developing SiloCRM and SiloCSM. My work spans the full product lifecycle: understanding business requirements, designing technical solutions, building features and integrations, deploying changes, and monitoring production systems.
A significant part of my recent work involves applied AI engineering:
- AI agents for voice calls, SMS, and email
- LLM integrations, RAG, and MCP-based tools
- Multi-tenant SaaS architecture
- Visual workflow automation engines
- Production infrastructure, CI/CD, monitoring, and backups
I also build independent products and open-source software. I care about making systems reliable, measurable, and useful in real-world workflows.
Currently interested in: Full-Stack Software Development, Applied AI, LLM Applications, SaaS Engineering, and Systems Integration.
SiloCRM & SiloCSM — JWeis Marketing
I work across the development and operation of multi-tenant SaaS products, including:
- Client and stakeholder requirement discovery and technical solution design.
- Full-stack product development across frontend, backend, and APIs.
- AI agents for inbound/outbound voice calls, SMS, and email.
- ElevenLabs voice-agent integration and Twilio communication services.
- RAG-based knowledge retrieval and per-tenant knowledge-base workflows.
- MCP integrations that expose CRM capabilities and data to AI agents.
- Visual workflow automation with 33 triggers, 20+ actions, conditional branching, and multi-step execution.
- Production deployments, CI/CD, infrastructure monitoring, backups, and operational optimization.
Most of this work lives in private company repositories. The public projects below demonstrate my independent engineering work.
|
LLM Evaluation Platform A platform for testing prompt quality and measuring LLM reliability.
Stack: Next.js · TypeScript · FastAPI · Python · PostgreSQL |
HVAC Field Service SaaS A multi-tenant platform designed for small HVAC teams.
Stack: Next.js · Fastify · Supabase · PostgreSQL · Drizzle |
|
QLoRA Research Investigating whether small fine-tuned language models can match BanglaBERT on noisy Bangla sentiment data.
Stack: Python · PyTorch · Hugging Face · QLoRA |
Open-Source Productivity Tool Turns a phone or tablet into a customizable Stream Deck for Windows.
Stack: Node.js · Preact · TypeScript |
|
Local Network File Sharing A Windows and Android application for chatting and sharing files over a local Wi-Fi network.
Stack: Flutter |
Chrome New Tab Experience A feature-rich new-tab page designed to improve everyday browsing and productivity. |
- Understand the problem first. Good software starts with understanding the actual business workflow.
- Design clear interfaces. API contracts and system boundaries matter as much as implementation details.
- Build for isolation. Multi-tenant systems need deliberate data-access and authorization boundaries.
- Measure AI quality. An LLM feature needs evaluation, regression checks, and observable execution.
- Think beyond deployment. Monitoring, backups, recovery, and maintainability are part of shipping software.
Autism Spectrum Disorder Identification from Facial Images Using Fine-Tuned Pre-trained Deep Learning Models and Explainable AI Techniques
Hossain, S. S., et al. (2025). Semarak International Journal of Applied Psychology, 5(1), 29–53. First author.
- Compared 12 fine-tuned CNN architectures.
- DenseNet121 achieved 90.33% accuracy and 90.00% F1-score.
- Used LIME and Grad-CAM for model interpretability.
I'm interested in working on challenging software problems involving full-stack development, SaaS platforms, AI applications, and system integrations.
If you're building something meaningful in these areas, I'd be happy to connect.
