SOFTWARE ENGINEER/ FULL-STACK · BACKEND · APPLIED AI

Rahul
Vignesh.

I turn problem statements into end-to-end solutions. Full-stack software, reliable APIs and AI workflows, built to work together.

Explore my work Discuss a role

GitHub LinkedIn View résumé

Open to software, FDE & AI roles/Sydney, Australia

LATENT SPACE / RV–01 DATA NETWORK

THE WHOLE SYSTEM.
FROM FIRST INPUT TO REAL IMPACT.

+RV / ENGINEER AT A GLANCE

From a problem to working software.
Here’s how I build.

I start with the problem: who needs the solution, how their workflow works and what a useful result looks like. Then I design and build across interfaces, APIs, data and AI, with testing, evaluation and deployment built into the process. I'm completing my studies at UNSW, graduating in November 2026, and looking for software, full-stack, backend, AI engineering, FDE and AI consulting roles across Australia.

Meet Rahul

01 / BACKEND & SEARCH

Answers with evidence.

BM25 and vector search fused with reciprocal ranks, then reranked with a cross-encoder. Cited generation, document permissions and a 50-case evaluation workflow make the pipeline inspectable.

02 / WORKFLOW & INTEGRATION

Models are one part of the job.

Gemini matching and tailoring in a prototype with persistent workflow state, human approval and audit logs, connected through FastAPI and React.

03 / FULL-STACK & DATA

From forecasts to decisions.

XGBoost inference, a 720-hour recursive forecast and optional PatchTST fusion, with tests for timestamps, lag updates and missing inputs. Served through FastAPI and explored in React.

+01 / SELECTED WORK

Three projects.
The engineering behind them.

The problem, the implementation and the hard decisions.
Full-stack software, data integration and applied AI.

FULL-STACK / BACKEND / APPLIED AI

01FULL-STACK / DATA / ML

EnergyAI

A full-stack application that combines building-demand forecasts, weather, meter history and solar estimates to explore energy-planning scenarios.

Engineering challengeKeeping 720 hours of recursive predictions aligned with changing lag inputs, and making missing-data fallbacks and historical adjustments explicit.

React
FastAPI
Python
XGBoost
PatchTST

View case study GitHub repositoryRESEARCH

Inside the build

XGBoost forecasts building demand from weather, calendar features and meter history. FastAPI serves inference and solar-planning APIs; React makes forecasts, historical adjustments and planning estimates explorable. An optional PatchTST path runs when its checkpoint and sufficient history are available.

Full-stack development and machine learning research

≈ 26.5MObservations

1,512Buildings

18Sites

HourlyObservations

BUILDING METADATAWEATHERHISTORICAL DEMAND

Temporal featuresXGBoostOptional PatchTSTRecursive forecasting

BDG-2 source dataset scale; not a training-row count

01 / Ingest

Load documents, split them into chunks and preserve source metadata for later citations.

02BACKEND / SEARCH / AI

Document RAG System

A document search and question-answering service that retrieves permitted source material, ranks relevant passages and returns cited answers or declines when evidence is insufficient.

Engineering challengeCombining keyword and semantic relevance while enforcing document access, then separating retrieval failures from unsupported answers during evaluation.

Python
RAG
ChromaDB
Groq
BM25

View case study GitHub repositoryIN DEVELOPMENT

Inside the build

An end-to-end RAG pipeline ingests documents, persists local embeddings in ChromaDB, combines BM25 and semantic retrieval, reranks the candidates and generates cited answers with Groq. Weak evidence triggers abstention.

Retrieval pipeline, API and application engineering

01 / Résumé ingestion

Extract candidate information from résumé documents and store relevant context for matching and later reuse.

03WORKFLOW / INTEGRATION / AUTOMATION

JobAgent AI

A full-stack prototype for comparing resumes with job requirements, drafting tailored material and tracking review steps through a persistent workflow.

Engineering challengeCoordinating model output, saved state and human approval without confusing a prepared or simulated action with a confirmed application.

Python
FastAPI
Gemini
RAG
ChromaDB

View case study GitHub repositoryIN DEVELOPMENT

Inside the build

React and FastAPI connect résumé ingestion, Gemini analysis, ChromaDB matching and SQLite application state. Matching and tailoring workflows include review steps and audit logs. Job discovery and Playwright application adapters currently simulate outcomes rather than submit real applications.

Agent orchestration and full-stack system development

Browse the project archive

+02 / THE SYSTEMS VIEW

I build the connections.
Across the whole application.

Good software engineering connects interfaces, services and data people can actually use.

SOFTWARE SYSTEM EXPLORERSELECT A CAPABILITY TO EXPLORE

EnergyAI: Smart Building Energy PlatformRESEARCH JobAgent AIIN DEVELOPMENT Document RAG SystemIN DEVELOPMENT

HOW I BUILD SOFTWARE

Every layer.
One considered system.

Start with the decision.

Understand the user’s actual problem, constraints and definition of success before choosing a model.

03 / THE PERSON BEHIND THE WORK

SYDNEY, AUSTRALIA / SOFTWARE ENGINEERING

I’m Rahul.
I build what I’m curious about.

I turn problem statements into working software. I work from requirements and system design through interfaces, APIs, data and AI workflows, then test and refine how the pieces work together.

Full-stackBackendApplied AI

Explore my work

01 / FOUNDATIONS / 2022–2024

It started
with robotics.

I spent my first two years at VIT Chennai studying Robotics and Artificial Intelligence. Working across software and physical systems shaped the way I approach engineering.

9.05VIT GPA

RoboticsArtificial intelligence

View my resume

02 / EDUCATION / 2024–PRESENT

Computer science.
Applied at UNSW.

I transferred to UNSW Sydney in 2024 to continue my studies in computer science and engineering. My honours research focuses on AI-based energy management in building operations and renewable energy.

Transfer studentHonours research

View my resume

03 / RESEARCH

Making energy
more predictable.

EnergyAI connects XGBoost inference to a 720-hour recursive forecast and optional PatchTST fusion. I work on lag updates, timestamp alignment and missing-input fallbacks, then expose the predictions through FastAPI and React.

Time seriesXGBoostEnergyAI

Explore EnergyAI

04 / PROJECT ENGINEERING

Beyond a
model response.

My document RAG pipeline fuses BM25 and vector search, reranks passages with a cross-encoder and generates cited answers. A 50-case evaluation workflow separates retrieval checks, answer quality and abstention, alongside document permissions and regression tests.

Hybrid retrievalRerankingLLM evaluation

Explore the RAG system

05 / CONTRIBUTION

Build together.
Learn in the open.

I contributed to open-source tools for video-based human movement analysis through the NUS OpenCap project. Working with an existing codebase is part of how I learn to make useful, considered changes.

OpenCapMovement analysisOpen source

View my resume

06 / THE NEXT CHAPTER

Useful software.
Built with a team.

I’m looking for software, full-stack, backend, forward-deployed and AI engineering roles across Australia. If your team is building a product that needs practical thinking, careful implementation and curiosity, I’d love to talk.

Software engineeringFDEApplied AI

Let’s connect

+04 / RESEARCH & EXPERIMENTS

A question.
Then a closer look.

Inside the lab

EXPERIMENT / TIME SERIES

Learning the rhythm of a building

Energy forecasting across hourly demand, weather and building context.

EXPERIMENT / MODEL FUSION

When models work together

XGBoost forecasts, optional PatchTST fusion and the effect of historical anchors on monthly energy estimates.

EXPERIMENT / RETRIEVAL SYSTEMS

Context before generation

Hybrid retrieval, reranking and citations for document question answering that can decline unsupported requests.

+05 / ENGINEERING SKILLS

Breadth across the stack.
Visible in the work.

From user interfaces and API integration to data, testing and applied AI. Each area connects to an implementation you can explore.

Full-stack development

React and TypeScript interfaces connected to Python APIs, with shared data contracts and clear loading, error and review states.

Python, TypeScript, JavaScript, React, Next.js, HTML/CSS

Explore the full-stack application

Backend & integration

FastAPI services that validate inputs, enforce document permissions and connect model, weather and data services. Familiarity with Model Context Protocol (MCP).

FastAPI, REST APIs, Pydantic, Authentication, Access controls, API integration, MCP (familiarity)

Inspect the service design

Data & persistence

Ingestion, relational workflow state and vector retrieval, with explicit handling of timestamps, lag features and incomplete history.

SQL, PostgreSQL, SQLite, ChromaDB, pandas, NumPy, Data validation

Explore data and workflow state

Applied AI & ML

Generative AI and agentic workflow prototypes, designed around retrieval, saved state, human review and evaluation. Also XGBoost forecasting and familiarity with multimodal AI.

Generative AI, Agentic AI, AI system design, RAG, BM25, Cross-encoders, Gemini, Groq, XGBoost, PyTorch, TensorFlow, Multimodal AI (familiarity)

See the retrieval implementation

Testing & reliability

Regression checks, typed contracts and structured logs, plus a 50-case evaluation workflow for retrieval, answer quality and abstention.

pytest, Regression testing, Debugging, Request logging, Ragas, LLM evaluation

See the validation approach

Implementation & delivery

Requirements analysis through implementation and deployment, using Codex and Claude Code with code review and verification. Docker packaging, CI checks and technical documentation support delivery.

Requirements analysis, Codex, Claude Code, Git, GitHub Actions, Docker, Vercel, Technical documentation

Explore a deployed application

+06 / FOUNDATIONS

Learning.
Applied.

Sydney, Australia

University of New South Wales

Computer Science Engineering studies

2024–Present. Transferred from VIT after the first two years of study. Thesis: AI-Based Energy Management in Building Operation: Integrating Renewable Energy.

Chennai, India

Vellore Institute of Technology (VIT)

Robotics and Artificial Intelligence, B.Tech studies

2022–2024. Completed the first two years at VIT Chennai before transferring to UNSW. GPA: 9.05.

+07 / A CONVERSATION STARTER

A different way
to explore the work.

Ask about the projects, the tools,
or how the pieces fit together.

Ask RahulGROUNDED IN THE PORTFOLIO

What would you like to know?

Portfolio search · source-linked answers · no invented achievements

+08 / WHAT’S NEXT?

Let’s build something
useful.

Software engineering · Applied AI · Collaboration

Start a conversation

OPENING THE SYSTEM

Loading the next page…

From a problem to working software.
Here’s how I build.

Meet Rahul

Beyond a
model response.

Hybrid retrievalRerankingLLM evaluation

Explore the RAG system

RahulVignesh.

From a problem to working software.Here’s how I build.

Answers with evidence.

Models are one part of the job.

From forecasts to decisions.

Three projects.The engineering behind them.

FULL-STACK / BACKEND / APPLIED AI

EnergyAI

Document RAG System

JobAgent AI

I build the connections.Across the whole application.

Every layer. One considered system.

Start with the decision.

I’m Rahul.I build what I’m curious about.

It startedwith robotics.

Computer science.Applied at UNSW.

Making energymore predictable.

Beyond amodel response.

Build together.Learn in the open.

Useful software.Built with a team.

A question.Then a closer look.

Learning the rhythm of a building

When models work together

Context before generation

Breadth across the stack.Visible in the work.

Full-stack development

Backend & integration

Data & persistence

Applied AI & ML

Testing & reliability

Implementation & delivery

Learning.Applied.

University of New South Wales

Computer Science Engineering studies

Vellore Institute of Technology (VIT)

Robotics and Artificial Intelligence, B.Tech studies

A different wayto explore the work.

Let’s build somethinguseful.

RahulVignesh.

From a problem to working software.Here’s how I build.

Answers with evidence.

Models are one part of the job.

From forecasts to decisions.

Three projects.The engineering behind them.

FULL-STACK / BACKEND / APPLIED AI

EnergyAI

Document RAG System

JobAgent AI

I build the connections.Across the whole application.

Every layer. One considered system.

Start with the decision.

I’m Rahul.I build what I’m curious about.

It startedwith robotics.

Computer science.Applied at UNSW.

Making energymore predictable.

Beyond amodel response.

Build together.Learn in the open.

Useful software.Built with a team.

A question.Then a closer look.

Learning the rhythm of a building

When models work together

Context before generation

Breadth across the stack.Visible in the work.

Full-stack development

Backend & integration

Data & persistence

Applied AI & ML

Testing & reliability

Implementation & delivery

Learning.Applied.

University of New South Wales

Computer Science Engineering studies

Vellore Institute of Technology (VIT)

Robotics and Artificial Intelligence, B.Tech studies

A different wayto explore the work.

Let’s build somethinguseful.

Rahul
Vignesh.

From a problem to working software.
Here’s how I build.

Three projects.
The engineering behind them.

I build the connections.
Across the whole application.

Every layer.
One considered system.

I’m Rahul.
I build what I’m curious about.

It started
with robotics.

Computer science.
Applied at UNSW.

Making energy
more predictable.

Beyond a
model response.

Build together.
Learn in the open.

Useful software.
Built with a team.

A question.
Then a closer look.

Breadth across the stack.
Visible in the work.

Learning.
Applied.

A different way
to explore the work.

Let’s build something
useful.

Rahul
Vignesh.

From a problem to working software.
Here’s how I build.

Three projects.
The engineering behind them.

I build the connections.
Across the whole application.

Every layer.
One considered system.

I’m Rahul.
I build what I’m curious about.

It started
with robotics.

Computer science.
Applied at UNSW.

Making energy
more predictable.

Beyond a
model response.

Build together.
Learn in the open.

Useful software.
Built with a team.

A question.
Then a closer look.

Breadth across the stack.
Visible in the work.

Learning.
Applied.

A different way
to explore the work.

Let’s build something
useful.