Work
AI Intern
AI Intern · Home Team Science and Technology Agency (HTX)
Robotics Process Automation and UIPath with xCloud; LangChain and LangGraph with xDigital
What I did
In xCloud, I was attached to the Enterprise-RPA team to learn more about UIPath and RPA. I built a few RPA bots for internal use.
I was more involved in xDigital, where I spearheaded efforts to adopt Agentic AI for Paperwork, an internal AI tool designed to streamline procurement processes. There, I:
- Compared agentic frameworks. I ran a comparative analysis and presented the findings to Paperwork’s core team to inform decision-making.
- Prototyped Paperwork v2.0 in LangGraph. I designed an autonomous state graph of agents with tool-calling capabilities that can extract and format key information from procurement documents. (Remember, this was back in 2025, when AI and agents were not as capable as they are today. Many of the things done back then are probably outdated by now)
- Optimised Llama4-Scout prompts. Using advanced prompt engineering techniques (chaining, meta-prompting), I improved response consistency by 15% in benchmark tests.
What I learned
- Being open minded! Initially, RPA wasn’t on my radar for my internship. I expected to come in to learn full stack, or even cloud computing (as the name xCloud suggested), so when I initially heard I was going to do RPA for my entire internship, I had a lot of doubts. But, getting exposure to it was quite useful for my subsequent projects (see my digitalisation project for my dad’s business), so I guess I managed to connect the dots backwards.
- Grabbing opportunities when I see them! A good friend of mine was already in HTX, and he made the jump from xCloud to xDigital. He told me what he was working on, and I felt it was a great learning opportunity as well. There was some hesitation, since I didn’t want to burn bridges with xCloud, but eventually I figured out an arrangement to be involved in both xDigital and xCloud, and I really did manage to learn a lot in xDigital, spearheading Agentic AI workflows, an aspect that the wider team has not looked at before.
- Presenting your work is as important as doing good work! I had a lot of ideas for the proof of concept, and we managed to make something work! My supervisor was determined for us to show the good work we did, and we spent a good deal of time in the last 2 weeks of our internship just working on slides, refining and rehearsing. I didn’t understand why at first, but after we presented the work, we realised how important it was to communicate technical ideas to non-technical members. I learnt a lot especially from a fellow intern, when he pulled out a slide titled Seeing Is Believing he managed to grab everyone’s attention, and everyone was visibly more impressed.