Which of my projects best demonstrates my suitability for this role?

Author: Mustafa Siddiqui · Part: 1 of 3 · Recruiter question: Which of Mustafa's projects best demonstrates his suitability for this role?

Evidence rule: This is my first-person authored source material. SudoChat may summarise it in third person but must not strengthen, exaggerate or invent claims. Limitations and Do not claim sections are hard factual boundaries.

# 20. Which of my projects best demonstrates my suitability for this role?

**Author:** Mustafa Siddiqui
**Source type:** First-person authored response
**Canonical recruiter question:** Which of Mustafa's projects best demonstrates his suitability for this role?

> This source is intentionally written in my first-person perspective. SudoChat should use it as evidence and answer external visitors in third person without strengthening, exaggerating, or removing the limitations recorded below.

## Direct answer
SudoChat is the project that most directly demonstrates my suitability for the Court AI Technologist role because it combines Copilot Studio, RAG, responsible AI, grounding, hallucination reduction, user experience and explicit governance boundaries in one practical proof of concept.

However, SudoChat is strongest when viewed alongside my broader delivery experience. SudoChat demonstrates how I think about responsible AI, while projects such as the DXC document approval automation, my government enterprise work and privacy preserving chatbot research demonstrate that I can apply those ideas to real organisational problems and controlled environments.

## Evidence
### SudoChat is the closest technical match to the role

SudoChat was deliberately built as more than a chatbot demonstration.

I wanted to explore how an AI assistant could operate when accuracy, evidence and limitations matter.

The system was therefore designed around:

* Copilot Studio
* retrieval augmented generation
* controlled knowledge sources
* grounding
* evidence based responses
* hallucination reduction
* explicit limitations
* "do not claim" boundaries
* safe fallback behaviour
* user experience
* responsible AI principles

The knowledge base itself was designed so the assistant does not simply receive positive information about me.

Each major entry includes:

**Direct answer**

**Evidence**

**Relevance**

**Limitations or gaps**

**Useful links**

**Do not claim**

This structure is intentional.

It gives the model information about both:

**what the evidence supports**

and

**where the evidence stops.**

That is directly relevant to responsible AI in a high trust environment.

### SudoChat demonstrates grounding rather than unrestricted generation

I chose a RAG based approach because the information the assistant needs is current, controlled and auditable.

Instead of relying on the LLM to remember facts about me, SudoChat retrieves relevant evidence when a question is asked.

Conceptually:

**User question**

↓

**retrieve relevant knowledge**

↓

**provide that evidence to the model**

↓

**generate a grounded response**

↓

**return the supporting information**

This is important because the system is designed around a principle I believe would also matter in the Federal Courts:

**authoritative information should remain separate from the language model wherever practical.**

### SudoChat demonstrates that refusal can be good AI behaviour

One of the most important design principles in SudoChat is that the assistant is not required to answer everything.

If someone asks:

**"What AI system did I deploy at the Federal Courts?"**

the correct response is not to infer an impressive answer from related experience.

It is to say that I have not deployed an AI system within the Federal Courts.

Similarly, if someone asks whether I led a government department's Copilot rollout, the assistant should distinguish between my actual contribution and the broader project.

This demonstrates my belief that:

**an AI system knowing when not to answer can be as important as knowing how to answer.**

That philosophy is particularly relevant to Court technology.

### SudoChat demonstrates responsible AI through architecture

The project incorporates several of my core responsible AI principles.

**Transparency**

Users should understand that they are interacting with an AI system and should be able to understand the source of important claims.

**Accountability**

The system should not create facts simply because the language model is capable of producing a plausible response.

**Safety and reliability**

Known failure modes such as hallucination and unsupported inference are considered deliberately.

**Privacy**

The wider architecture raises questions about what information an AI assistant actually needs access to and where that information should be processed.

**Human control**

SudoChat is an assistant rather than an autonomous decision maker.

### SudoChat demonstrates user centred design

I also built the experience around the user rather than leaving Copilot Studio as a default chat interface.

The custom SudoChat experience includes:

* a purpose built frontend
* suggested questions
* clear navigation
* an Engine Room explaining the system
* a chat focused layout
* visible explanation of the architecture
* limitations and responsible AI messaging

The Engine Room is particularly relevant.

I wanted users to be able to see that behind a seemingly simple chatbot is a wider architecture involving retrieval, knowledge, instructions and controls.

This reflects my belief that AI systems should not be presented as unexplained magic.

Knowledge Base · Next part →