Software · Applied AI · Adelaide

Agrim Sharma

Final-year Computer Science student at Adelaide University, majoring in Artificial Intelligence. I build software that solves problems, put AI to work inside them, and ship it so real people actually use it.

Available for internships now · graduate roles from December 2026

agrimsh22@gmail.com

Projects

Problem, approach, and impact. The stack comes second.

Pacific Village Explorer

Top 12 finalist · 2026 Humanitarian Innovation Hackathon

Problem

A village council deciding where to rebuild after a flood, where to move a school, or how to farm ground turning saline has no way to see its own coastline in twenty years. The digital twin platforms that answer that cost over $10,000 per site and need technical staff villages do not have.

Approach

Built in 44 hours with a three-university team at the University of Sydney. A coordinator picks their village, drags a timeline from 2026 to 2075, and watches houses, wells, farms and sacred sites change status as the water rises, with adaptation options for every asset at risk.

I led the AI recommendation engine. The prompt carries each village's real context, its population, cultural constraints, budget and community capacity, into the Claude API, with guardrails that keep every recommendation tied to that village instead of generic climate advice, plus an offline fallback so the tool still works with no API at all.

The call I am proudest of was 2D over a 3D digital twin. 86% of the Pacific has mobile coverage but only 27% use mobile internet, so the real constraint is usability, not connectivity. Appropriate technology beats impressive technology.

Impact

Top 12 of 75 submissions from 110 teams, and the only finalist representing Adelaide University. Judged by a panel including Sir Peter Cosgrove, Canva, Commonwealth Bank's AI Scientist and the Pacific Region Infrastructure Facility. Under $500 per village against $10,000+ for the alternatives, on any phone from the last five years, built on IPCC AR6, SPREP and NASA sea level data. The Grand Final is on 19 August.

Source Finalist announcement

React · Vite · Tailwind · Leaflet · Claude API · Vercel

Agrim Sharma with his two teammates at the Humanitarian Innovation Hackathon — The team

Whiteboarding the app architecture during the hackathon — Whiteboarding the build

The three-university team building during the 44-hour hackathon — 44 hours in

MetaPlay

Live · 2025

Problem

Gamers track what they play across scattered notes, spreadsheets, and memory. Our brief was a single place to discover games, build a collection, and review them, backed by real game data rather than a static seed list.

Approach

I worked in a five-person team and owned the front-end and back-end integration. I built the authentication layer, with email and password login plus Google sign-in across three roles (admin, user, and guest), and designed a normalised MySQL schema across eight tables.

After the course ended I came back to it on my own, fixed the remaining issues end to end, wired it up to live RAWG game data, and deployed it to production.

Impact

It runs live today with personalised dashboards, collections, reviews, and an admin panel, pulling real-time data from the RAWG API. This is the deployed version of a team project that I took the rest of the way and shipped myself.

Live demo Source

Vue.js · Node.js · Express · MySQL · Passport.js

Python · 2026

Pathfinder AI Agent

A logic-based agent that reasons its way through hidden hazards. It scored ten out of ten on every hidden autograder map.

It builds a propositional knowledge base from what it senses and deduces which cells are safe, taking a calculated risk only when forced. My clearest example of logic-based AI and reasoning under uncertainty.

Python · Propositional Logic · Search

Source

Java · 2026

GPS Tracker Dashboard

A real-time tracking dashboard built on functional reactive streams, verified by nineteen unit tests.

Location updates and alerts flow through composable event streams instead of shared mutable state, so the UI updates cleanly as data arrives. It is where I learned to handle streaming data and event-driven design properly.

Java · Sodium FRP · JUnit

Source

C++ · 2025

Virtual Restaurant Simulator

A restaurant simulation designed around clean class hierarchies, so new staff roles extend it without rewrites.

A Person base class with inheritance and virtual methods lets each role define its own behaviour through a shared interface, and state persists to files between sessions. Adding a staff type means adding a class, not rewriting logic.

C++ · OOP · Makefile

Source

Leadership

Adelaide University Tennis Club · President, Aug 2024 to Mar 2026

A dormant club to Club of the Year in eighteen months.

I took over a club that had gone quiet and rebuilt it from scratch. I led the merger of two university tennis clubs during the Adelaide and UniSA consolidation, co-authored the new constitution, brought together an eight-member committee, and won grants for facility upgrades. The real challenge was getting two groups who didn't know each other to trust the process and work as one.

Adelaide University Sport Club of the Year cheque presentation — Club of the Year · Adelaide University Sport, 2025

2 → 1
clubs merged: $7,000+
grants secured: 10 → 100+
active members: Winner
Club of the Year 2025

The challengeRunning our side of a two-university club mergerRead it

Setup

Adelaide and UniSA were merging, so every affiliated club had to merge too. I was president of the club that was ahead on membership and results, so I ran our side of it. That meant recruiting and bringing the right people across, aligning two committees that each ran their own social and competitive tennis, leading the merged constitution, and keeping every stakeholder comfortable, the University most of all, with as few conflicts as possible.

What I did

I started by working through what each committee actually wanted and where the two overlapped, then led the merged constitution until both sides were happy to sign it. The hardest piece was the coaches. They were two independent businesses, each loyal to a different club, and any shared model had to close the gaps where money could quietly leak out of the club. We talked through several models over a lot of meetings and settled on a three-month trial with both coaches running sessions together, so the real problems would surface early and we could fix them before locking anything in. Alongside that I ran member feedback to keep people bought into the change, and took on the practical handover myself: the facilities, the nets and fencing, the spare rackets, and the full asset register.

Result

One merged club running smoothly, both coaches retained on a model that actually worked, the members kept happy through the transition, and Club of the Year in the same year.

Experience

Where I've led and worked

Across leading a committee, coaching more than fifty clients of every age and background, and working a busy retail floor, the common thread is working with people. It's made me good at reading and communicating with people from all walks of life, and comfortable fitting into any environment, including the Australian workplace I want to build my career in.

Aug 2024 to Mar 2026

President, Adelaide University Tennis Club

Led the club revival covered above: merger, constitution, committee, grants, and the award.

Mar 2024 to Present

Tennis Coach, Tea Tree Gully Tennis Club

I coach juniors and adults at all skill levels, both one-on-one and in groups. I plan each session, adjust my feedback to the player, and track their progress over time. It has made me good at explaining the same idea a few different ways until it clicks, and at keeping people motivated when they are finding it hard.

Feb 2024 to Sep 2024

Retail Assistant, IGA Supermarkets

Part-time customer service across two stores while studying full-time. Busy retail floor, steady standards, and a lot of practice juggling work and study at the same time.

Mentorship

Career Access Mentoring Program · Adelaide University

Credential

A structured 14-hour mentorship through Adelaide University, spread across meetings over four to six months, mentored one-on-one by an APT Analyst at Google Threat Intelligence. We have stayed in touch for more than a year since. He ran me through a full mock Google interview loop, from the online assessment to the technical, behavioural and HR rounds, and was direct about exactly where I was weak and what to work on. He showed me how each round actually works, how to structure an answer, and which of my own experiences to draw on for it. We also covered how engineers at that level approach software craft and problem-solving, and where my strengths fit best. We met in person for the first time at Google's Sydney office, where he gave me a tour, and he still reviews my thinking whenever I am making a decision about my career.

University of Adelaide Certificate of Completion for the Career Access Mentoring Program

Visiting my mentor at Google Sydney

Agrim Sharma beside the neon Google sign in the Google Sydney office — Google Sydney office

Certifications

Technology Job Simulation · Deloitte AustraliaCredential2026
Data Analytics Job Simulation · Deloitte AustraliaCredential2026
Partnering with AI in the Workplace · DatacomCredential2026

About

I build the way I play tennis, which is to say I care about the boring part. Fifteen years on court teaches you that the good shots come out of a lot of unglamorous, repeatable work, and that following through when it stops being fun is most of the game. I still coach, and watching a player finally land a shot on the fortieth try feels a lot like watching a stubborn build go green.

Leading the tennis club turnaround is where that became how I think about software. Getting a group of people who didn't agree at first onto one plan is not far off getting a tangled codebase into a state the next person can actually pick up. Both come down to follow-through, and to still caring once the interesting part is over.

When a humanitarian AI hackathon came up in Sydney, I backed myself, booked the trip interstate on my own, and built for 44 hours with people I met that weekend. We finished in the top 12. Unfamiliar rooms on short deadlines turn out to be where I do some of my best work.

Agrim Sharma playing a forehand — Match play · Adelaide

A few things I hold to, that my projects show

I would rather ship something real people use than perfect a prototype nobody sees.

MetaPlay began as a group assignment. I took it the rest of the way on my own and put it in production, so today it is a live product instead of a repo I describe in interviews.

I build so the next person can extend it without a rewrite.

The Restaurant Simulator was designed around exactly that: adding a new kind of staff means adding a class, not editing the code that already works.

I reach for the simplest thing that solves the real problem.

Not the most impressive one. The point is the person on the other end, not the framework I get to use.

Right now I am building two tools I want in my own hands, a voice log for tennis coaches and a receipt scanner for Australian GST, studying for the Azure AZ-900, and getting ready for the hackathon Grand Final in August.

AI major
Computer Science: Dec 2026
Graduating: 15 years
On court: Adelaide
Based in Australia

Contact

I'm looking for graduate and internship roles in software engineering, AI, data, and analytics. Email is the best way to reach me. I'm based in Adelaide and happy to relocate within Australia.

Email me

Currently in Australia on a student visa with work rights for internship and part-time roles. Full, unrestricted working rights from 15 December 2026 (graduation), eligible for the subclass 485 Temporary Graduate visa. Available for graduate roles from December 2026, and for internships immediately.