Sydney, Australia · Founder of Vargos Labs

Vadi Taslim

Product & Systems Engineer

Software · AI · Infrastructure · Interactive Systems

Get in touch Résumé

VT

Resume

vlimz77@gmail.com

Vadi Taslim

Product & Systems Engineer | Sydney, Australia

Get in touch

Experience

September 2026 - Present

Founder & Product Engineer at Vargos Labs

▹Founded an AI-assisted engineering initiative exploring how individuals and small teams can build and operate businesses with fewer resources.
▹Designed and operated an agentic development lifecycle combining AI agents, automation, software integrations and human oversight.
▹Building practical solutions across property services, education and emerging-market businesses, including generative AI customer support and operational workflows.

Show 1 more...

▹Own product discovery, architecture, implementation, infrastructure and ongoing operation, with a focus on accessible, low-cost software solutions.

TypeScript·Next.js·Node.js·PostgreSQL·MCP·LLM Agents

July 2021 - September 2026

Senior Software Engineer at Vervio Digital

▹Led frontend architecture and delivery across global React, Node.js, TypeScript and Next.js applications, including migrations from legacy applications to modern Next.js architecture.
▹Led development and global rollout for Global Kitchen Appliance's cart and checkout experiences across multiple regions, integrating CommerceTools, and payment platforms including Adyen, PayPal, and Klarna.
▹Reduced application bundle size by approximately 50% through refactoring, dependency splitting, caching strategies and compiler-level optimisation.

Show 4 more...

▹Delivered content-driven customer experiences integrated with Adobe Experience Manager (AEM), headless CMS platforms, Algolia, product information systems and internal APIs.
▹Built supporting backend workflows using AWS (primarily Lambda, Step Functions, API Gateway, SQS/SNS and CloudWatch).
▹Built a mobile-first, SEO-friendly and WCAG-compliant Next.js application from the ground up and led a five-person team through the transition.
▹Established and maintained engineering practices around automated testing, accessibility, code quality, application performance and maintainable frontend architecture.

Next.js·React·TypeScript·Node.js·AEM·Algolia

August 2020 - February 2022

Software Engineer at Toptal

▹Built production React, Node.js, TypeScript and Next.js applications from scratch while working directly with distributed US product teams and business owners.
▹Owned frontend architecture and supporting Node.js services, including applications built with React, Redux, TypeScript, PostgreSQL and Redis.
▹Engineered server-side rendering for an existing React/Redux application using Express and Webpack before later migrating the application to Next.js.

Show 2 more...

▹Delivered payment integrations, CI/CD workflows, Docker deployments, CMS integrations, automated testing and application migrations.
▹Built responsive, mobile-first, SEO-friendly and WCAG-compliant user experiences across multiple client products.

React·TypeScript·Redux·Next.js·Node.js·PostgreSQL

Earlier experience (2012–2020)

Software EngineeratWallethub2018–2020

▹Developed and maintained large-scale fintech applications across JavaScript, TypeScript, Angular, Node.js, PHP, Redis, SOLR and MySQL.
▹Took ownership of OWASP ZAP automation and developed security-scanning workflows for common web vulnerabilities and known CVEs.
▹Investigated exploits and vulnerabilities escalated from security reports submitted through Bugcrowd and HackerOne.

Show 1 more...

▹Improved responsive interfaces, automated testing, SEO and application performance across production systems.

Angular·TypeScript·Node.js·PHP·OWASP ZAP

Frontend EngineeratAwe2017–2019

▹Owned core WebGL and Three.js rendering components for interactive 3D web applications, including VR, AR and XR capabilities.
▹Learned and applied matrices, trigonometry, geometry and linear algebra to engineer core 3D rendering functionality.
▹Developed and shipped interactive Three.js components and plugins while working around evolving browser and WebGL capabilities.

Show 1 more...

▹Benchmarked browser scripting, rendering and painting performance and diagnosed memory leaks and performance bottlenecks.

Three.js·WebGL·WebXR·JavaScript

Frontend EngineeratDominopos2016–2017

▹Developed modern responsive applications across Angular, Node.js, SCSS and MySQL.
▹Introduced unit-testing practices to existing projects from the ground up and worked with QA to integrate testing into the development lifecycle.
▹Reviewed and modernised fragile or outdated implementations while establishing consistent mobile-first frontend practices.

Show 1 more...

▹Won the company's annual CodeCamp competition for rapidly building a modern five-page web application from scratch.

Angular·Node.js·SCSS·MySQL

Fullstack EngineeratLogistics Consulting Asia2015–2016

▹Developed full-stack logistics applications and worked directly with design and management teams to translate operational workflows into software.
▹Proposed and implemented scalable replacements for deprecated code, improving application performance and maintainability.
▹Architected backend functionality with unit and functional test coverage to reduce regressions across existing workflows.

Show 1 more...

▹Built and maintained custom client reporting systems and implemented enhancements based on operational requirements.

PHP·Symfony·PHPUnit·Selenium

Fullstack EngineeratPowerGEN2012–2015

▹Architected and built a large-scale Windows desktop application for facial recognition and fingerprint identification using .NET, DigitalPersona and Neurotechnology APIs.
▹Developed and maintained frontend, backend and desktop applications for a diverse set of clients.
▹Worked directly with clients and management to translate business requirements into technical solutions.

Show 1 more...

▹Provided production support and performed root-cause analysis for live application issues.

C#·.NET·MSSQL·DigitalPersona·Neurotechnology

Projects

I build what I can't stop thinking about - usually because I want the thing to exist and nobody else has built it yet.

Vargos.ai

Businesses don't want a tool, they want the problem solved. Personal AI agents also disappear after each message and can't reach your systems.

Vargos Labs: software, AI and people combined to deliver business outcomes. Built on an open-source, self-hosted agent OS with persistent memory, messaging channels (WhatsApp, Telegram), scheduled autonomy, real tool access, and Agentic Development Lifecycle (ADLC).

Learnings:

▹LLM orchestration at system level
▹Agent autonomy vs control tradeoffs

TypeScriptMCPPostgreSQLpgvectorSQLiteWebSocketPlaywrightRAGVoice AI

CramClass

Families preparing for selective school exams want tutors who sat the exams recently, not generic AI or generic tutoring.

An online tutoring platform where high-ATAR university students share live selective-school and scholarship exam experience with families. Built by Vargos Labs from zero to MVP in 4 weeks.

Learnings:

▹Scoping a client MVP down to a 4-week launch
▹Wiring booking, payments and calendar into one flow

Google MeetGoogle CalendarStripeGoogle CloudAI Automation

Chans.ai

Slapping voice onto an AI agent means cobbling together STT, TTS, and WebRTC yourself. That sucks.

Drop real-time voice into any LLM. It's just an API - STT, TTS, voice chat, all connected to your model.

Learnings:

▹Real-time audio streaming with LiveKit and WebRTC
▹Hybrid RAG search (vector + BM25) with line-level citations

PythonFastAPILiveKitNext.jspgvector

Archived

Ingra.ai

A place to dump functions and let LLMs call them as tools.

Tic Tac Toe

A tic-tac-toe AI you can't beat because it's using minimax.

URL Shortener

A quick link shortener so I don't have to copy/paste long URLs between my phone and desktop.

Writing

View all

Writing about the decisions behind the systems - why something was built that way, and what went wrong first.

ai·Oct 2026·5 min

Qwen3.8 Flash-Next on a 32 GB R9700: 2 Users at 250K Context

Flash-Next Gyro-S on 1 R9700 with 32 GB of VM RAM: 2 users passed a 249K-token retrieval test, while 4 hit a runtime watchdog.

ai·Aug 2026·5 min

Two Rigs, Five Runtimes: My Qwen3.8-27B Benchmark Across NVIDIA and AMD

Qwen3.8-27B across two rigs and five runtimes: NVFP4 on Blackwell hits 150 tok/s, the AMD box peaks at 57 tok/s.

ai·May 2026·9 min

Qwen 3.6-35B-A3B on an RTX PRO 6000: 217 tok/s at Full bf16

Speed, latency, context scaling, and quality notes for Qwen 3.6-35B-A3B running at full bfloat16 on a single RTX PRO 6000 Blackwell via vLLM.

About

Currently based in Sydney. Started in game design, then into software engineering. Over the years I've built biometric identity platforms, 3D renderers, and global commerce systems that ran across dozens of countries. Game design taught me how to empathize with users.

Now I run Vargos Labs, a Sydney company that takes a business problem and delivers the outcome, starting with software. Behind it: multiple compute and GPU servers, fine-tuned local models, and Vargos, an open-source agent OS with real tool access on my infrastructure. Self-host because the cloud hides the bugs I care about.

Technologies

Frontend

Fast, SEO-friendly interfaces that handle real users

TypeScriptPrimary language

React10+ years

Next.jsApp Router, SSR, ISR

Tailwind CSSDesign systems

Backend

APIs that don't fall over, data pipelines, automation

Node.jsProduction services

PythonAI/ML, automation

PostgreSQLPrimary datastore

RedisCaching, queues

Infrastructure

Hardware I own, operate, and repair

DockerContainer orchestration

ProxmoxVMs, LXCs, GPU passthrough

CloudflareDNS, tunnels, edge

LinuxDaily driver

AI & ML

Running models locally without selling my soul to the cloud

vLLMHigh-throughput inference

OllamaLocal LLM serving

llama.cppQuantized models

WhisperSpeech-to-text

Stable DiffusionImage generation

LiveKitReal-time voice AI

Infrastructure

Cloud abstractions hide how systems actually behave. I self-host to see what gets hidden.

Intentional boundaries: Compete with hyperscalers · Replace managed services at scale · Build infra without operational purpose

What self-hosting taught me

▹Every abstraction leaks; knowing when matters more than avoiding them
▹Cold starts, IO contention, and memory pressure define real performance
▹Don't pick GPU specs until you know your actual workload
▹VRAM is the real GPU bottleneck - not compute
▹The bug is almost always in the network. It just looks like your app.
▹Disk and memory constraints surface before CPU saturation
▹Backups are useless until restores are tested
▹Observability beats intuition
▹Infra that isn't automated will eventually drift

Mobile dev environment

Mobile terminal

Termux (Android)cloudflaredtmuxClaude CLI

Full development environment from a phone - tunneled into my server, persistent sessions, agentic coding with Claude.

AI agent via chat

Telegram / WhatsApp / TermuxVargos (voice STT + text)Home server

Message my AI agent from any channel - text or voice - to execute tasks directly on my infrastructure.

Details

Machines

HYDRA

Primary Workstation · Windows 11 + WSL2

Intel Core i9-12900K

64GB DDR5 4400MHz

NVIDIA RTX 5080 (16GB)

2TB NVMe + 2x1TB NVMe

▹All-purpose development workstation
▹Local ML experimentation
▹GPU-bound workloads and tooling

HYDRA09

Secondary Server Node · Proxmox VE

AMD Ryzen 9 7950X

128GB DDR5

2x AMD Radeon AI PRO R9700 (64GB, ROCm)

2x2TB NVMe + 4TB NVMe + 2x18TB HDD RAID 1 ZFS-backed bulk storage

▹Virtualization (vargos-109, truenas, vervio, qantas)
▹AMD/ROCm GPU workloads (llama.cpp, vLLM, Stable Diffusion)
▹Secondary AI/ML Proxmox host
▹ZFS-backed bulk storage and NAS services

HYDRA10

Primary Server Node · Proxmox VE

AMD Ryzen 9 9950X

192GB DDR5

NVIDIA RTX PRO 6000 Max-Q (96GB)

512GB NVMe boot + bulk

▹Virtualization (vargos-110)
▹NVIDIA/CUDA LLM and Stable Diffusion inference workloads
▹Primary AI/ML Proxmox host and playground
▹GPU-bound workloads and tooling
▹Fine-tuning and model experimentation

Beelink SER9 Pro

Edge / Utility Node · Proxmox VE

AMD Ryzen 9-class mobile CPU

32GB DDR5

1TB NVMe

▹Edge services
▹Lightweight containers
▹Auxiliary workloads and testing

LEGION-S7

Mobile Control & Development Node · Ubuntu

AMD Ryzen 7 5800H

32GB DDR4

NVIDIA RTX 3060 (Laptop)

1TB NVMe

▹Primary mobile workstation for development

Networking

MikroTik RouterOS

▹Subnet and bridge isolation
▹NAT and firewall rule ownership
▹WireGuard VPN for secure remote access
▹Cloudflare-based DNS and ingress control

Platforms & Tooling

AI & ML Tooling

Local LLM runtimes (vLLM, llama.cpp, Ollama)Model experimentation and benchmarkingGPU utilization and performance analysis

Virtualization & Containers

Proxmox VE (VMs and LXCs)PCIe device passthrough (GPU, storage, networking)DockerDocker ComposeContainerized services on Proxmox LXCs

Networking & Access

MikroTik RouterOSWireGuard VPNCloudflare DNS and Tunnel-based ingressReverse proxy and internal routing

Data & Storage

ZFS (datasets, snapshots)NFS mounts between nodesNAS-backed persistent volumesPostgreSQLRedisMariaDBVector databases (Qdrant)

Media & Real-Time

Plex media streamingPhotoPrism photo indexingLiveKit real-time audio/video

Observability & Ops

Service health monitoringLog inspection and debuggingBackup orchestration and restore testing

Get in touch

Tell me what you're building.

vlimz77@gmail.com

Developer Q&A reputation

Top 3% of vetted developers

GitHub contribution streak statistics for Vadi Taslim

Open source contribution streak