wGrow · Singapore · Est. 2008
We are an eighteen-year-old Singapore software studio, now organised around AI agents. We put agent crews into the kinds of systems regulators inspect, with the same encryption, audit logs and PDPA discipline as before, and we deliver roughly ten times faster.
Claude Code · Cursor · custom evals · vLLM · .NET · Postgres
New: CTO Services
A fractional CTO and the engineers to execute, for one monthly fee. Led by Timothy Mo and Scott Li.
Explore CTO ServicesTrack record
What's different in 2026
Engineering has moved from writing code to specifying intent and reviewing the output. We rebuilt the studio's tooling around that.
In January, IMDA published the world's first governance framework for agentic AI. Buyers will start asking about it, and we already build to it.
Most AI agencies are about eighteen months old. We have held production for eighteen years, and the unglamorous parts of that work add up.
Who we work with
Most of our work this year falls into three groups. Each has its own procurement style and needs a different team, so the proposal you receive is written for your situation.
Existing wGrow clients
We can add an agent crew to your current .NET, MS SQL and AWS estate without rewriting it. Delivery speeds up and you keep the same audit trail.
Read moreSingapore government and regulated
We build inside the Model Governance Framework for Agentic AI launched at WEF 2026. PDPA controls, audit logs and role-scoped autonomy are designed in from the start.
Read moreDeep-tech and APAC scale-ups
Our embedded delivery team runs Claude Code and custom evals against your domain. We do this every week for WaterDoctor, an NUS spinoff.
Read moreWhat we sell
The agentic build is our main offering. Around it we provide infrastructure, training and coaching, and a private model on your own hardware for teams that can't put PDPA-class data on a public LLM endpoint.
Product engineering, websites, and SEO/GEO content from the Article Studio crew. We use the same .NET, Node, Postgres, MS SQL, AWS and Azure stack as before, now with AI integrated by default.
Read moreCIS-benchmark hardening for Linux and Windows, plus a secure AI routing bridge so engineers can use agentic CLIs without your data leaving the building.
Read moreHands-on training on Claude Code, Codex CLI and Gemini CLI. OpenClaw and Hermes Agents setup. Eval harness and context coaching for teams that have stalled after the demo.
Read moreOpen-weights LLMs on hardware you own: GPU procurement, vLLM or TGI deployment, optional LoRA finetuning, and an OpenAI-compatible internal endpoint behind your SSO and audit log.
Read moreAgent crews
One agent on its own rarely gets past the demo. A small team of agents with a human editor and an eval harness can run as a system. These three crews are running in our studio today, including the one embedded with a deep-tech client.
Lead research, outreach and proposal drafting
Reads inbound enquiries, checks them against ACRA, LinkedIn and tender feeds, drafts a tailored response and hands it to a person within the hour. The pipeline keeps running overnight.
Research, drafting, review and publishing
Produces most of the articles on this site, with a senior engineer as editor. Every claim has to trace back to a wGrow project or a cited source.
Live engagement, embedded with WaterDoctor
Four agents and two senior engineers embedded with WaterDoctor, an NUS spinoff building AI-integrated biofilm reactors. They turn sensor data into adaptive control loops and reports a regulator can read.
Flagship engagement
WaterDoctor is an NUS spinoff building AI-integrated biofilm reactors. As its strategic technology partner, wGrow designed and operates the agent crew that runs its adaptive treatment loops. Sensor and lab telemetry go in, and run reports a regulator can read come out.
Read the case study
Eighteen years of delivery
A short list. The rebrand is recent, but the foundation is older than most of the agencies announcing themselves this quarter.
AWS, MS SQL, field-level encryption, PDPA
10-step case workflow, auto billing, client portal
Vitasg, Xpressflower, PayNow and PayLah
Lazada, Shopee, TikTok, Amazon, unified inventory
Application portal, expert review, claims, audit reports
SS 564, IoT sensor ingest, PUE reporting
Sales and production balancing, safe-stock guards
Encrypted client data, annual invoicing, commission ledgers
QR attendance, SMS reminders, billing
Featured long-form
Our Article Crew produces most of these, with a senior engineer as editor. We try to publish what we learned that week rather than what's trending.
Paul Werbos derived backpropagation in 1974 on a machine that could not run it. The fifty-year wait between the math arriving and the compute catching up is the story of how 1970s ideas, not 2010s ones, built modern AI.
Massive AI context windows degrade reasoning and introduce silent code regressions when treated as infinite storage. Engineers must cap working context at 30 percent and enforce strict session hygiene using handover notes to maintain architectural continuity.
Massive context windows aren't working memory—they are unsearchable junk drawers that degrade agent reasoning. To prevent context collapse, multi-agent systems require a dedicated Curator agent to actively filter, deduplicate, and synthesize information.
Latest field notes
Technical write-ups, agent team updates and project notes, updated weekly.
Brief us
Tell us what you need. A wGrow engineer will reply within one business day.
Brief us on your problem [email protected]
