Dr Raminderpal Singh

From OpenClaw to project fred

Most agent frameworks — LangChain/LangGraph, LlamaIndex, AutoGen, CrewAI — are general-purpose toolkits. They stay neutral about the operating model, expecting the developer to compose primitives into whatever shape an application needs. OpenClaw is the opposite: narrow, opinionated, and purpose-built for one operating model — long-running, scheduled, parallel-thread research agents that gather information continuously, hold session state across runs, and produce reasoned output a scientist reviews. Every design choice falls out of that single operating model. For scientific R&D workflows, specialization beats generality: a general framework adapted to this shape carries permanent engineering tax — defaults that don't fit, abstractions that leak, hooks for use cases that don't apply.

My judgment that this operating model is correct is grounded in two decades of direct experience inside scientific R&D — spanning experimental design, data quality, regulatory considerations, and the recurring causes of pilot failure — combined with the work of architecting a full secure stack with OpenClaw at its core. ScienceClaw is the first production deployment of that stack, with ALS hypothesis generation, compound deep-dives for drug discovery, life science market intelligence, and scientific research assistants now under active delivery on the same foundation. Each engagement has shaped a precise understanding of where the operating model is correct, where the runtime defaults require adjustment, and how the security, orchestration, and observability layers around it should be designed.

Evolving the stack: project fred

I am now evolving this work into project fred — a secure-by-design, clean-room codebase that will, over time, replace OpenClaw at the core of my stack. project fred carries forward the operating model proven above and adds two capabilities the current runtime does not provide. First, the security architecture is being designed from the first commit, with requirements being shaped in active collaboration with academic and enterprise partners — rather than retrofitted onto an existing codebase. Second, and more fundamentally, the core engine is tuned to the behavior of the LLM driving it. Reasoning models behave differently from instruction-tuned models; cloud-scale models behave differently from local mid-size and local smaller models. An orchestration core that treats the LLM as an opaque substrate — as most frameworks, including OpenClaw, do — leaves significant control on the table. project fred treats LLM behavior as a first-class architectural concern.

The architecture is a core engine with three model wrappers — cloud LLMs, local mid-size LLMs, and local smaller LLMs, all thinking models — each with its own tuned interaction strategy. The core selects and coordinates across them according to the work being done and the deployment constraints in force.

project fred is being built from a nanobot starting point that already encodes the strengths of the OpenClaw operating model proven above — allowing the build to begin from a working baseline rather than from scratch, while every subsequent line of code lands in a clean-room repository under my control. This combination preserves what the four years of OpenClaw work have established about the right shape, and lets the new effort focus where it matters: the security architecture, the LLM-tuned core, and the deployment characteristics that academic and enterprise partners require for the next generation of scientific R&D agents.

Engineering discipline meets AI momentum

I'm an AI Solution Architect with a deep-rooted foundation in systems engineering — trained to understand how complex pieces fit together, where the friction is, what the boundaries between components should be, and what it takes to make something work reliably at scale. That instinct shapes everything I build.

What excites me about this moment is the raw momentum that AI brings. The speed at which ideas can become working software has fundamentally changed. But momentum without structure produces fragile systems, hallucinated outputs, and untestable code.

AI has made Systems Engineering more critical, not less. When AI writes the code, the architect is still accountable. The value a human brings is no longer in the implementation — it is in requirements elicitation, defining module boundaries, and deciding what the system should actually do. Those are systems engineering decisions that no AI agent makes reliably without that discipline enforcing the structure.

I particularly enjoy building software appliances for R&D scientists — purpose-built tools that run on OpenClaw to deliver autonomous, reliable research capabilities. ScienceClaw is the embodiment of this: an autonomous research platform that scientists can actually trust, built on rigorous engineering principles.

That's the thread running through all my work — from teaching scientists to vibe‑code responsibly, to building test‑first workflows that keep AI agents honest, to deploying autonomous research platforms. Two decades inside scientific R&D — understanding how experiments are designed, where data breaks down, what regulators care about, and why adoption stalls — means I build for the constraints that actually exist, not the ones that look good in a pitch deck.