Saakshi Gupta

|

AI Engineer focused on LLM agents, RAG and multimodal AI. I build scalable services and multi-agent architectures that automate workflows and deliver measurable impact.

Adelaide, Australia
// FEATURED PROJECTLIVE BETA

AlgoCanvas

An interactive algorithm visualization canvas. Step through sorting, graph traversal, and data-structure operations frame-by-frame with synchronized code highlighting and live narration.

bubble_sort.py
Frame 1/33

Select Workspace

Complexity Info
AlgorithmBubble Sort
Time ComplexityO(N²)
Space ComplexityO(1)
VIEW: CANVAS SURFACE_01
250
121
452
83
304
Synchronized DebuglogPython 3
1
def bubble_sort(arr):
2
    n = len(arr)
3
    for i in range(n):
4
        for j in range(0, n - i - 1):
5
            if arr[j] > arr[j + 1]:
6
                # Swap elements
7
                arr[j], arr[j+1] = arr[j+1], arr[j]
Narrator Execution Feed

Initial Array loaded. Click 'Play' or 'Step' to start Bubble Sort.

9 Visualizer Surfaces

Live, editable cards for arrays, graphs, trees, heaps, matrices, hash maps, linked lists, tries, and strings — each with its own bespoke renderer.

111+ Visualizations

Ready-to-run traces across 15 categories — sorting, shortest paths, tree ops, graph analysis, DP, strings, clustering, and more.

Drawing Canvas & I/O

Sketch structures with pen and shape tools, import PDF / PPTX / images, export the canvas to PNG / SVG, and jump anywhere with a ⌘K command palette.

// ABOUT

About me

Saakshi Gupta

Saakshi Gupta

Adelaide, Australia

I'm an AI engineer focused on LLM agents, RAG, and multimodal AI. I build scalable services and multi-agent architectures that automate workflows and deliver measurable impact — both in research and in production.

Most recently at CSIRO's Data61, I created a multimodal deepfake analysis system integrating CNN/Vision Transformer detection, SHAP/LIME for explainability, image-to-text captioning with CLIP, and an LLM contextual explanation module using Llama 3.1 and Ollama. The work was published as DF-P2E at ACM Multimedia 2025.

Before CSIRO I interned at LINQIA in San Francisco, where I led development of algorithms that analysed millions of creator profiles and matched their content styles to brand-campaign requirements. I also worked on multimodal video analysis using Gemini Vision and AWS Nova.

I contribute to open source frameworks I use day-to-day — Pydantic AI, LangChain, NVIDIA Megatron-LM, DeepAgents, Haystack, Matplotlib, MCP Atlassian — and write about agentic architectures, GraphRAG, and code visualisation on Medium.

// EXPERIENCE

Where I've worked

Four roles across AI engineering, data analysis, and research.

// LEADERSHIP & ACTIVITIES

Beyond the job experience

Selected leadership roles, community contributions, and recognitions from university and industry.

ACM-W, University of Queensland

Vice Chair

Promoted computing education and supported women in computing at UQ.

TEDxUQ

Web Developer & Leadership STV

Built intuitive interactions and visual designs for the TEDxUQ website.

Queensland AI Hub

Student Ambassador

Advocated for AI education and connected students with industry across Queensland's AI community.

University of Queensland

Student Partnership Projects

Developed interactive modules for advanced ML courses (COMP4702/COMP7703) and cyber-safety modules for international students.

UQ EAIT

Leadership Pathway Program

Completed intensive leadership development program with industry panels and workshops.

UQ Ventures

San Francisco Startup Adventure

Selected from 500+ applicants for a fully sponsored entrepreneurship immersion. Collaborated with startups on strategic challenges.

Westpac Hackathon 2024

Hackathon Finalist — QuickSettle

Built a group-payments app with shareable links and automated reminders at the Gold Coast Engineering Hub.

// PROJECTS

Things I've built

From production-grade multi-agent systems and explainable deepfake detection to fine-tuned LLMs and full-stack platforms. Click any card for a case study.

// SKILLS

Technologies & tools

What I reach for. Grouped roughly by problem space rather than alphabetised. Click any skill for a short blurb on how I use it and a docs link.

Languages

Day-to-day for shipping production code

AI & LLM Frameworks

Stack for building agents, RAG, and fine-tuned models

ML & Computer Vision

Training, inference, and explainability

Data & Infrastructure

Storage, observability, and orchestration

Backend & APIs

Services I reach for first

Frontend

Where the AI meets the user

Tools & Workflow

Daily drivers around the code

Currently learning

What I'm digging into right now

// DETAILED BLOG

Deep Technical Writing

Architecture maps, design tradeoffs, and implementation details from projects I build end-to-end. Plus shorter pieces published on Medium.

All posts
Also publishing on Medium — switch the toggle or see all posts →

// OPEN SOURCE

Contributions

10 merged pull requests across 8 open-source projects — mostly bug fixes and small features in the frameworks I use day-to-day: Pydantic AI, LangChain, NVIDIA Megatron-LM, Matplotlib, and others.

// GET IN TOUCH

Let's build something.

Saakshi Gupta

Saakshi Gupta

AI Engineer

Adelaide, Australia

Want to chat? Pick a time that works for you

30-min via Calendly

// SEND ME A MESSAGE