Verify, then trust
Independent critics, approval gates, heuristic fallbacks. A system is not correct because it ran.
Graduating December 2026Adelaide, SA
AdhinJeeva
Final-year Software Engineering (Honours) student building secure, reliable systems.
I design cloud-native backends, multi-agent AI pipelines and automation workflows — and I build them to be verified, not assumed. Networks & cybersecurity foundations, production habits, Adelaide-based.
About01 / 05
I build systemsthat get checkedbefore they get trusted.
I am a final-year Software Engineering (Honours) student at Flinders University, graduating in December 2026. My work sits where cloud infrastructure, API integration and applied AI meet — and the thread running through all of it is verification.
In my multi-agent LLM pipeline, every output routes through three critic agents backed by different providers, specifically so no single model’s blind spot goes unchecked. In the automation workflow I designed at the City of Salisbury, an AI drafts the synthesis but a human approval gate stands between it and anything final. In CivicPulse, credentials live in a secrets layer rather than in source.
None of that is accidental. I would rather build the guardrail than explain the incident, and I like working in the structure that makes guardrails stick — sprint cycles, code review, CI, written architecture.
- 12
- person Agile team led as System Architect
- 3
- critic agents, independently backed, in the QA pipeline
- 5
- automated workflow categories routed in CivicPulse
- ~80%
- estimated reduction in manual processing, CivicPulse
Secure by default
OAuth2 flows and secrets-based credential management over hard-coded keys, from the first commit.
Degrade gracefully
Fallback routines so an upstream API failure bends the system instead of breaking it.
Architecture is communication
As system architect on a 12-person Agile team, the diagram and the review were the deliverable.
Certifications
- AWS APAC Solutions Architecture Job SimulationForage · 2025
- Google Data Analytics Professional CertificateGoogle
- Generative AI for Software DevelopersIBM
Selected work02 / 05
Four systems,built to hold.
012025
CivicPulse
Cloud-Native Full-Stack Web Application
A cloud-native microservices application that reads inbound civic correspondence, classifies it with real-time AI and routes it automatically across five workflow categories.
- Role
- Architect & Developer
- Outcome
- 5 routing categories automated
- Outcome
- ~80% estimated manual processing removed
- Node-RED
- Node.js
- IBM Watson NLU
- Gmail API (OAuth2)
- AWS Lambda
- JavaScript
- +1
022025
Multi-Agent LLM Arbitration
Quality-Assurance Pipeline
A production-grade multi-agent system that arbitrates LLM output quality by routing every response through three specialised critic agents, each backed by a different provider so shared blind spots cannot survive.
- Role
- Sole Architect & Engineer
- Outcome
- 3 independent critic agents
- Outcome
- Async fan-out / fan-in orchestration
- Python
- LangGraph
- FastAPI
- Streamlit
- Google Gemini API
- Ollama (Llama 3)
- +4
03Current
Polaris Automation Suite
Workflow Automation & Research Tooling
Automation built inside the City of Salisbury’s Polaris Business Development Centre: multi-stage mentoring report generation with a human in the loop, plus a research toolchain for profiling local businesses.
- Role
- Business Development & Automation Support
- Outcome
- 1 human approval gate before any report ships
- Outcome
- 2 research skills shipped, single + batch
- Power Automate
- AI Builder
- Python
- Crawlee
- BeautifulSoup
- Apify
- +1
042025
Fringe Booking Platform
Team Capstone — Agile Delivery
A booking platform delivered by a 12-person Agile team, where my remit was the architecture itself: system design, delivery pipeline, and the review discipline that held the sprints together.
- Role
- System Architect
- Outcome
- 12 engineers on the delivery team
- Outcome
- CI/CD pipelines authored in GitHub Actions
- GitHub Actions
- CI/CD
- Agile/Scrum
- Jira
- Git
- Code Review
Skills & expertise03 / 05
The toolsI actually use.
Five groupings, pulled straight from shipped work and completed coursework. Hover or tap a category to read the full set.
Cloud & DevOps
- 01AWS EC2
- 02AWS S3
- 03AWS Lambda
- 04Docker
- 05GitHub Actions
- 06CI/CD pipelines
Back-End & APIs
- 01Node.js
- 02FastAPI
- 03PHP
- 04RESTful API design
- 05Microservices
- 06OAuth2
Databases
- 01MySQL (3NF design)
- 02PostgreSQL
- 03SQLite
- 04Firebase
- 05SQL querying
Tools & Process
- 01Git
- 02GitHub (PRs, review, branching)
- 03Jira
- 04Agile/Scrum
- 05V-Model
Languages
- 01Java
- 02Python
- 03JavaScript
- 04TypeScript
- 05PHP
- 06C
- 07SQL
- 08Go
Every item listed is drawn from work I have shipped or coursework I have completed — no self-assessed percentages, no proficiency bars.
Path04 / 05
How I gothere.
Current
Experience
Business Development & Automation Support
City of Salisbury — Polaris Business Development Centre
- Designed a Power Automate + AI Builder workflow automating multi-stage mentoring report generation, with an AI-drafted synthesis step and a human approval gate before reports are finalised.
- Built Claude Code skills integrated with Apify actors for automated B2B company research: a single-company profiling skill and a batch leads-finder across Salisbury and Adelaide.
- Wrote a custom Python web-scraping actor using Crawlee and BeautifulSoup.
- Power Automate
- AI Builder
- Python
- Apify
- Crawlee
2025
Project
Multi-Agent LLM Arbitration & QA Pipeline
Independent project
- Architected and delivered a production-grade multi-agent system on LangGraph, routing output through three critic agents on separate LLM providers.
- Built an asynchronous FastAPI backend with fan-out/fan-in orchestration and heuristic fallback routines for API resilience.
- LangGraph
- FastAPI
- Gemini
- Ollama
- Docker
2025
Project
CivicPulse — Cloud-Native Full-Stack Application
Project
- Architected and deployed a cloud-native microservices web app with automated routing across five workflow categories.
- Integrated Gmail API over OAuth2 and IBM Watson NLU for real-time classification; implemented secrets-based credential management.
- Node-RED
- AWS Lambda
- OAuth2
- IBM Watson NLU
Expected Dec 2026
Education
Bachelor of Engineering (Software) (Honours)
Flinders University, Adelaide SA
Key coursework
- Networks & Cybersecurity · Data Structures & Algorithms · Software Testing & QA
- Cloud & Distributed Computing · Databases · AI Fundamentals
- Software Engineering: People, Process & DevOps
- Networks & Cybersecurity
- Software Testing & QA
- Cloud & Distributed Computing
Contact05 / 05
Let’s buildsomething solid.
Graduating December 2026 and looking for graduate software engineering and cyber-adjacent work in Adelaide. If you are building something that has to be right, I would like to hear about it.
- jeevaadhin@gmail.com
- in/adhin-jeeva
- GitHub
- adhinjeev
- Based in
- Adelaide, SA