Start here
AI Engineer
Hi, I'm Ashir. I specialise in building AI systems by turning prototypes into production-ready applications.
Now at Latham & WatkinsStack Python · C# .NET · TypeScriptCloud Microsoft Azure
About Me
Identify. Build. Ship.
I'm an AI & Automation Engineer with software engineering roots. I build and ship AI systems to production, from AI-powered applications and agentic RAG to multi-agent orchestration, across cloud and full-stack environments.
I do my best work where innovation and ownership are both expected. I pick things up fast, I go looking for the complicated problems, and I tend to leave things in better shape than the brief asked for.
SpecialismAI SystemsAgent orchestrationAgentic RAG
CertifiedAZ-900AI-900
My Projects
Things I build outside work
View all projects4 more in the archive, including ENbe Education
Experience
Where I've worked
Nov 2025 - current
AI & Automation Engineer
Latham & WatkinsA standalone role owning end-to-end discovery and delivery of AI use cases for customer support, automating processes across ServiceNow, Talkdesk and Azure.
- Built a custom summarisation and guidance skill: retrieval over past records and knowledge articles, scripts to standardise LLM input, and prompt generation for agent collaboration. It brought clarity, structure and direction to the Workplace Experience Team.
- Architected a multi-agent chatbot workflow, now the first response for technology support: RAG over knowledge articles for basic troubleshooting, diagnostic workflows that find performance bottlenecks on user devices, and service portal equipment ordering.
- Deployed and maintained multiple AI agents that autocomplete tickets by analysing similar past tickets.
Nov 2024 - Nov 2025
Software Engineer
CapgeminiKey developer on the AI Sales Assistant, an award-winning .NET (Blazor) web application powered by agentic RAG on Azure AI Services. It retrieves internal client knowledge to accelerate bid creation, and supported over £250m of successful sales activity in 2025.
- Shipped user-facing features including multi-file selection in conversational flows, multi-modal input handling, model-specific configuration controls and a system admin conversational view — improving usability, flexibility and governance for internal stakeholders.
- Led a full function schema redesign to fix GPT-4.1 tool invocation issues introduced by multi-modal workflows, pushing tool call reliability and response quality past the original implementation.
- Improved maintainability and resilience by identifying and implementing architectural best practices and defensive programming patterns that cut debugging time and increased production stability.
- Optimised Azure OpenAI response streaming - error handling through asynchronous programming, plus UI responsiveness fixes for prerendering conflicts that reduced latency.
Hackathons
1st placeCapgemini Global Microsoft / NVIDIA / GitHub Agentic AI Hackathon
1,420+ participants · multi-agent virtual lab built on MCP servers1st placeCapgemini Global Microsoft / GitHub AI Reinvention Hackathon
Automated client reference creation, Azure + Semantic KernelSkills
What I am known for
Programming languages
PythonC# (.NET)TypeScript
Technologies & frameworks
Next.jsBlazor .NETVue.jsDjangoSemantic KernelLangChain
Cloud & DevOps
Azure AI FoundryAzure App ServicesAzure DevOps Pipelines
Education
Where I studied
2021 - 2024
BSc (Hons) Computer Science - 2:1
University of Leeds- Final year project: GymPal, an AI-driven gym assistant built by fine-tuning and quantising pre-trained LLMs on custom datasets curated through web scraping. Focused on conversational flow and model performance under tight resource constraints.
Algorithms and Data StructuresDistributed SystemsArtificial IntelligenceMachine Learning
Contact Me
End of the line
Have anything you want to build together, or problems you want to solve? I am always open to new opportunities and collaborations, so feel free to reach out to me via the form below or through any of my social links.