Nucleusbox – ML, Agent Engineering & Open Source
(ML & AI Atheneum – Be More. Be Fluid.)
Nucleusbox is where applied machine learning meets agent engineering: learn with clarity, build with structure, and ship systems that survive beyond the demo.
Our philosophy
We believe learning and production are one journey, not two websites.
Most people meet ML through tutorials. Most teams meet agents through demos. Both break the same way: the first win feels easy; the second challenge exposes missing structure.
Nucleusbox exists to close that gap.
| Principle | What does it mean here |
|---|---|
| Be more | Go deep enough to understand why, not just copy a notebook. |
| Be fluid | Models, tools, and providers change. Your thinking should adapt. |
| Demo ≠ done | A working prototype is the start. Maintainability is the product. |
| Open by default | Blog, newsletters, dev tools, and NucleusIQ are built to be shared. |
The biggest risk is not building the first demo. It is surviving the next twelve months of learning, maintenance, and real use.
That line guides our blog, our Agent Engineering newsletter, and our open-source work.
What we do
Three things, one mission: help engineers grow from curiosity to production.
1. Learn: Applied ML & AI
Our Get Started path and blog cover Python for data science, ML fundamentals, deep learning, and modern AI. Written for practitioners who want a clear roadmap, not hype.
Start where you are. Stay consistent. Two focused hours a day beats scattered weekends.
2. Build: NucleusIQ (open source)
NucleusIQ is our agent-first Python framework for AI agents that survive production: long runs, heavy tool use, real stakes.
Every agent works in a demo. Almost none survive production without structure. NucleusIQ is the harness:
- 3 execution modes: Direct, Standard, Autonomous (gearbox, not one-size-fits-all)
- Context management before the model breaks
- Provider portability: OpenAI, Gemini, Anthropic, Groq, Ollama, MCP
- Telemetry on every run: know what happened, not guess
pip install nucleusiq nucleusiq-openai→ GitHub · Docs · PyPI · Philosophy
3. Use: Free developer tools
Browser-based utilities for ML and AI engineers: JSON, YAML, token counting, diff, regex, and more. Private by design. Runs locally in your browser.
The Nucleusbox ↔ NucleusIQ connection
Nucleusbox is the home: education, community, and applied engineering content.
NucleusIQ is the open-source framework we build and dogfood. The same ideas we write about (context, guardrails, evals, harness design) encoded as software.
We do not sell a black-box platform. We teach, we open-source, and we show our work.
Team
Brijesh Kumar Singh, Founder, Nucleusbox
Applied ML & AI engineer with 12+ years in data, including 5+ years at Informatica. MSc in Machine Learning & AI from Liverpool John Moores University.
I write the Agent Engineering series, maintain NucleusIQ, and publish practical guides on this site for engineers who want to learn deeply and ship reliably.
Work with me
- Freelance (data & AI): Fiverr: birju1100
- Open source: GitHub: nucleusbox
- Questions on NucleusIQ: GitHub Discussions
- Blog & tools: you’re already here.
For collaboration, consulting, or speaking, reach out via Fiverr or GitHub. We read every discussion thread on NucleusIQ.
Join the work
- ⭐ Star NucleusIQ if production agents matter to you
- 📬 Follow the newsletter: Agent Engineering on LinkedIn
- 🛠 Try the developer tools (free, no upload to our servers)
- 📖 Start learning: Get Started with Applied ML
Open source
NucleusIQ: agents that survive production
Agent-first Python framework: 3 execution modes, context engine, MCP tools, telemetry on every run.
pip install nucleusiq nucleusiq-openai
