[HIRING] OpenClaw + VPS person — paid trial task, ongoing work | Remote, any timezone

Jobs
1

I run a signage manufacturing business in Australia. Over the last while I’ve built an AI agent stack — OpenClaw agents on a hardened self-hosted VPS, with n8n coming in to handle the deterministic stuff (tracking signs through our production line, Telegram photo capture, monday.com dashboards). One production box is live and healthy. Now I’m scaling to more instances and I’ve hit the limit of doing all the technical work myself. I need a proper technical partner for the long run.

**First job — a paid trial task, fixed price:**

I’ll give you a fresh throwaway VPS (Hetzner, Ubuntu 24.04). Turn it into a hardened OpenClaw host:

- OpenClaw installed properly — pinned version, local npm, dedicated non-sudo user, running under systemd

- Hardened: SSH key-only, UFW, fail2ban, everything loopback-bound except :22

- Monitoring: CrowdSec + Netdata — or make the case for whatever you’d use instead. I care more about your reasoning than the exact tools.

- Secrets in env files, 0600, never in unit files or shell history

- Hand back: the working box, a short runbook of what you did and why, and a screen recording of the important parts (Loom or whatever — voiceover is great, no camera needed). I archive these so we can repeat and reference everything later.

If the trial goes well there’s plenty of ongoing work — more agent boxes, n8n workflows for the production line, and on-call fixes when something breaks. I’m after someone with serious hours in this space, who’s patient enough to explain things step by step. Screen recording the significant work is standard on every job here, not just the trial — that’s how I keep it all documented.

**To apply, reply with these four things:**

1. **Answer this one:** a staff member sends a photo of a finished sign to our Telegram agent. In 3–5 sentences, how would you get that photo stored and linked to the right job record? No single right answer — I want to see how you think.

2. One link to the most relevant thing you’ve built. Just one, the closest match — not your whole portfolio.

3. Your fixed quote for the trial + hourly rate for ongoing work.

4. Your timezone and how fast you can usually jump on something urgent.

Fair warning: if your reply doesn’t answer question 1, I won’t read the rest. I’m getting a lot of copy-paste pitches lately and I’d much rather a short honest reply from a real practitioner than a polished brochure.

And even if you’re not the hire this round — I’m building a network of good automation people and I keep names. Happy to talk to plenty of you.

2 Likes
2

Hey :waving_hand:,

I’m Milan, with **8 years** of experience in Business Automation and AI. Including 2 years at **Apify** working on enterprise-level browser automation.

Currently specializing in n8n, but also proficient in Python & Javascript.

Find out more about my work here:

* **[Youtube Channel](https://youtube.com/@vasarmilan/)**

* **[Smoothwork.AI Website](https://smoothwork.ai)**

* **[Public N8N Workflows](vasarmilan | n8n Creator)**

If you think I might be a match, please:

**[Book a call here](Book a Call - SmoothWork)** with me

Or reach out at **[hello@smoothwork.ai](mailto:hello@smoothwork.ai)**

Looking forward to hearing from you!

3

If you need I am available here

I am one of top 50 verified template creators:

I am also capable of building custom community n8n node too:

  1. https://www.npmjs.com/package/n8n-nodes-pdfbro
  2. https://www.npmjs.com/package/n8n-nodes-ocrbro
  3. https://www.npmjs.com/package/n8n-nodes-ttsbro

Apart from that I’m also a full-stack developer with the right Gen AI experience, which makes me a solid plus for your team[but right now only vibecoding]

Check my recent gen ai projects… I built a native Android automation agent too. It’s worth a look:

I can build complex AI automations directly in code, not just inside n8n

I recently started posting my n8n work on YouTube with explanations:

You can schedule a quick call with me: N8N Project Consultation with Vaar | Iamvaar | Cal.com

Fun fact… I even made an n8n workflow to find a few n8n project leads for myself so I truly believe in what I do…

9bSkKu
9bSkKu1220×1470 122 KB

I started asking My recent clients for honest feedbacks so here is one testimonial: https://www.youtube.com/watch?v=TqBy3SVCHgQ&list=PLAJltY5bp6yiZ3sFBjm7bfrkLXSGtJX8m

Here is my linktree: iamvaar Official: X | Linktree

And I am also capable of building gohighlevel automations and other CRM’s too, here are my recent gohighlevel integrated automation videos:

And I also built low latency voice appointment scheduler with live ai avatar[in code]:

I built AI search visibility tracker[I am capable of building complex web scraping automations in python too]:

And done many other big projects too which are under NDA’s so technically I can’t reveal them.

4

Hello @James_Nation would love to work with you on this, you can book a call here Calendly to discuss the requirements and also you can check out my Upwork profile Upwork

5

James, for the photo path: the Telegram agent should capture the file id, staff user, timestamp and any job number in the caption. The file gets copied out of Telegram into a stable storage path keyed by job id, not left as a Telegram-only asset. n8n can do the deterministic part: validate the job in monday.com, write the photo URL plus Telegram message id back to the job record, then send the staff member a short confirmation. If the job match is ambiguous, it should go to a review queue instead of guessing.

Closest public thing: AI Workflow Reliability Proof | n8n, LLM API, CRM Handoff — synthetic, but it shows the intake → controlled automation → logs/runbook style I use.

Fixed trial quote: USD 520 for the Ubuntu/OpenClaw host, hardening, systemd service, env handling, monitoring choice, runbook and screen recording. Ongoing work: USD 70/hour.

Timezone: UTC-3. For urgent break/fix, I can usually pick it up within 2-4 hours during my working window and same day outside that; for anything that needs pager-style coverage, the response window should be agreed up front.

1 Like
6

Hi @James_Nation,

Here’s my response to your application requirements.

1. Telegram Photo → Job Record Workflow

When a staff member sends a finished-sign photo to Telegram, I’d have Telegram trigger an n8n webhook containing the photo, caption, sender details, and timestamp. The workflow would identify the correct job using a job number in the caption, QR/barcode data, or production context already stored in Monday.com. Once matched, the image would be stored against the relevant job record (either uploaded directly to Monday.com or archived in object storage with the link written back to the job). I’d also record metadata such as who submitted it, when it was received, and optionally move the job to the next production stage automatically.

2. Most Relevant Work

My background is primarily in automation systems, AI agent orchestration, workflow engineering, and production-grade infrastructure. One of my recent projects involved building an AI-driven interview platform and large-scale automation pipelines using n8n, self-hosted services, and custom integrations.

3. Pricing

For the trial task, I’m flexible on pricing and would prefer to align on a fair fixed scope once I’ve reviewed the exact requirements and expectations for the build, documentation, and handover.

For ongoing work, I typically work on either an hourly or milestone basis depending on the nature of the project. My focus is on building long-term partnerships rather than optimizing for a one-off engagement, so I’m happy to structure something that works well for both sides.

4. Timezone & Availability

  • I’m comfortable working around Australian business hours and have no issue aligning my schedule when required for deployments, troubleshooting, or project discussions.
  • For production-impacting issues, I can typically respond within 30–60 minutes during working hours and within a few hours outside business hours.

Best regards,

Paul

7

Hi James,

  1. For the Telegram photo path, I would not let the agent guess the job record. The Telegram step should capture file_id, sender, timestamp and caption/context, then n8n should validate a stable job key against monday.com. The image should be copied out of Telegram into a durable job-keyed storage path, with the file URL, Telegram message id, sender and capture timestamp written back to the monday.com item. If the job match is missing or ambiguous, it should go to a review queue and ask the staff member for a job number instead of updating the wrong record.

  2. Closest public thing I can share: https://taxinoromadrid.es - not a VPS-hardening portfolio page, but a recent end-to-end operational web project. The VPS/agent/runbook work I do is usually internal rather than public.

  3. Fixed quote for the trial: USD 600. That includes a fresh Ubuntu 24.04 host, dedicated non-sudo user, pinned OpenClaw install, local npm/runtime, systemd service, SSH key-only login, UFW, fail2ban, loopback-bound services except SSH, env files with 0600 permissions, CrowdSec + Netdata unless we agree on a better fit, update/rollback notes, short runbook and screen recording. Ongoing: USD 70/hour, or fixed price per box/workflow where scope is clear.

  4. Timezone: Europe/Madrid (UTC+2). I can usually pick up urgent break/fix within 2-4 hours during my working day/evening, and for production-impacting work we can agree specific response windows before I become responsible for it.

I would start once the fixed trial is agreed and funded, only on the throwaway VPS, with no production credentials, and keep the handoff notes clear enough that you can repeat the setup later.

Carlos

8

Hi James,

Interesting setup, and welcome to the community.

For the Telegram photo workflow, I’d have the Telegram bot receive the image and pass the file ID plus metadata into n8n. The workflow would identify the correct job either from a job reference included in the message or by matching against active production records, then store the image in the chosen storage layer and write the image URL back to the corresponding monday.com job record. I’d also log timestamps and user details so every photo remains traceable throughout the production lifecycle.

I’ve sent you a DM with a few additional details.

Folafoluwa

9

Hi James, I can take a shot at the paid trial.

For the Telegram photo flow, I would not let the bot guess the job. I’d make the staff either select the active job, reply inside a job thread, or include/scan the job number before upload. n8n would keep the Telegram message_id/file_id, download the image, store it in the agreed storage, then update the right monday.com item with the file link plus a small photo log: staff, timestamp, source message, checksum. If the match is unclear, it goes into a review queue instead of being attached to the wrong job. I’d also make the upload idempotent so the same photo cannot create duplicate records.

Closest public sample: OpenClaw VPS trial runbook sample · GitHub

Trial quote: US$380 fixed. Ongoing: US$35/hr.
Timezone: UTC+8. I can start today and overlap with Australia evenings. For agreed urgent windows I can usually jump in within 2-4 hours.

10

Hey James,

1. Photo-to-job-record flow:

The Telegram bot receives the photo via webhook. I’d extract the chat ID and timestamp from the incoming message, then use a two-step match: first, check if the sender’s Telegram user ID maps to a specific job (e.g., staff are assigned to active jobs in a lookup table in Airtable or a simple JSON config). If the sender has only one active job, auto-link it. If multiple, prompt them with quick-reply buttons listing their active job IDs. The photo gets saved to cloud storage (S3/Backblaze) with a naming convention like {job_id}_{timestamp}.jpg, then the storage URL is written back to the job record in monday.com via API. The whole flow runs in n8n: Telegram trigger then lookup then save then monday.com update, with a fallback to a human if the match is ambiguous.

2. Most relevant build:

I built an AI-powered outbound sales agent using n8n + Vapi AI + Airtable - automated cold calling with real-time product recommendations from a live catalog, CRM pipeline updates, and follow-up logic based on call outcomes. Full n8n workflow with modular nodes for easy extension.

3. Quote:

Trial task (hardened VPS setup): $200 fixed. Ongoing: $40/hr.

4. Timezone and availability:

UTC+8 (Beijing). I can typically respond to urgent issues within 30 minutes during waking hours (9am-midnight local), and within 2 hours overnight. I am used to async-first communication with Australian and European teams.

Best,
Xiaoyu

11

Hi James,

Your setup is exactly the kind of practical automation work I like: agent boxes for real operations, n8n for deterministic flows, and documentation that makes the system repeatable instead of tribal knowledge.

  1. For the Telegram finished-sign photo flow, I would have the Telegram bot/n8n trigger capture the message metadata, file_id, sender, timestamp, and any job reference in the caption or conversation context. n8n would fetch the actual file through the Telegram Bot API, store it in object storage or a controlled folder using a path like jobs/{job_id}/finished-signs/{timestamp}.jpg, then update the matching monday.com job record with the image URL/attachment and an audit note. If the job cannot be matched confidently, I would route it to a small review queue instead of guessing, then use that correction to improve the matching rule.
  2. Closest relevant work: recent hands-on work around Codex-assisted AI workflow orchestration, VPS/proxy setup, agent-style SOPs, and Feishu/Lark task/documentation hubs. I can share a short runbook or screen recording as part of the trial, which matches the way you archive repeatable work.
  3. Fixed quote for trial: USD $120 for one fresh VPS setup within the stated scope, plus a short handoff runbook and screen recording. Ongoing rate: USD $30/hour. If the scope expands beyond one fresh VPS box, I would quote the delta before doing extra work.
  4. Timezone: China, UTC+8. I can usually respond async within the same day, and for urgent break/fix I can usually jump in within 2-6 hours depending on local time.

My working style is to keep secrets out of shell history/unit files, document every meaningful decision, and leave behind a box that someone else can reason about later.

12

Hi James,

On the photo → Monday job record question:

The system extracts whatever the photo shows — job number, customer name, sign type — through a vision LLM, then cross-references Monday’s active “sign installed” records via the GraphQL API (so it never breaks when Monday updates their UI). One high-confidence match (~85%+), it auto-links. Borderline confidence or multiple records could match, it lands in a review queue — one-click confirm on your side, and the system gets sharper on the edge cases every time you confirm.

The failure mode I’d want to never happen: a wrong-record auto-link you don’t catch for weeks. The 85% threshold + human gate makes that structurally impossible — not “unlikely,” impossible.

On the build: your spec (UFW, fail2ban, SSH key-only, loopback-bound services, CrowdSec + Netdata, env files at 0600) is exactly the production hygiene I run for my own self-hosted work. ClaudeBot, my Claude Agent SDK daemon, lives on the same baseline — append-only audit log, tool-security gates on every action, persistent memory across restarts. Walking that through Loom is the trial deliverable you scoped.

Quote:
• Trial: $450 fixed — working box + runbook + Loom recording
• Ongoing: $95/hr

Why $95/hr — security hardening that holds under production load is the work most automation builders skip past. I’d rather be priced as the person who does it right than the one you replace in three months.

Timezone: IST (UTC+5:30). 10 AM my morning = 2:30 PM your afternoon — clean overlap for daily syncs, async for the rest.

Can start the trial within 24 hours of go-ahead.

Portfolio: priyanshukumar.co

Best,
Priyanshu Kumar

13

Dear Client,

I am writing in response to your OpenClaw and VPS infrastructure opportunity. After reviewing your setup description, I can see this is not a basic deployment task but a production oriented automation stack with real operational dependency across OpenClaw agents, n8n workflows, Telegram ingestion, and monday.com based tracking. That level of system integration and reliability focus is exactly the type of environment I work best in.

I also understand the importance of security hardening, repeatability, and clear documentation through runbooks and screen recordings. Your approach of treating infrastructure as a repeatable system rather than one off setups is aligned with how I structure automation and backend deployments.

Below are my responses to your required questions.

Answer to Question 1

When a staff member sends a photo of a finished sign to the Telegram agent, I would first capture the incoming message through a Telegram webhook connected to a processing service running on the VPS. The image file would be downloaded immediately and stored in a structured object storage system with a unique identifier.

Next, I would extract or map metadata from the Telegram message such as sender ID, timestamp, and any embedded job reference or caption text. This metadata would be used to query the production tracking system, likely through n8n, to match the image to the correct job record in monday.com or the internal database.

Once the job is identified, I would update the corresponding record with the image URL and metadata, ensuring traceability between production stage and final output. Finally, I would log the event in a structured audit log so that every image ingestion can be traced back for quality control and operational review.

Relevant Experience

I have worked on backend automation systems involving API integrations, webhook driven workflows, and AI assisted task routing using tools such as n8n, custom Node.js services, and cloud based VPS deployments. My focus has been on building reliable pipelines where data flows between messaging platforms, databases, and operational dashboards without manual intervention.

Trial Quote and Ongoing Rate

For the initial VPS hardening and OpenClaw setup trial task, my fixed quote would be 450 USD.

For ongoing infrastructure and automation work including agent scaling, workflow development, and maintenance, my hourly rate would be 35 USD per hour.

Timezone and Availability

I am based in India Standard Time and typically available within 1 to 3 hours for urgent issues during active working windows. For critical production issues, I can usually respond faster depending on severity.

I am comfortable working in environments that require careful documentation, repeatable infrastructure setup, and iterative scaling of automation systems. I would be interested in contributing to the long term development of your agent based production workflow and ensuring it remains stable as it scales.

Thank you for your time and consideration.

14

Hey @James_Nation

1. Telegram photo → job record: The staff member’s Telegram message hits an n8n webhook. I’d parse the caption or reply-context for a job ID (or prompt them via the bot if missing), pull the photo via Telegram’s getFile API, push it to object storage (S3/Backblaze or monday.com file column directly), then update the matching monday.com item — attaching the file, stamping the status column, and logging timestamp + sender. If the job ID is ambiguous, the bot replies asking for confirmation rather than guessing. Keeps it deterministic, auditable, and reversible.

2. Closest match: Monday Setup & Automation | Loom (monday.com setup & automation — pipeline/dashboard work closest to your production-tracking use case)

3. Pricing:

  • Trial (hardened OpenClaw VPS + runbook + screen recording):
  • Ongoing hourly: $35/hr

4. Timezone & response: IST (Mumbai) — roughly 4.5 hr overlap with AEDT mornings. For urgent issues I’m typically on within 30–60 min during my working day, and reachable on Telegram/WhatsApp outside it for genuine fires.

Happy to jump on a quick call if the above lands. I’d rather under-promise on the hardening side and walk you through my exact approach to OpenClaw pinning, systemd unit, and secrets handling before we kick off, than oversell it here.

15

I run self-hosted VPS boxes (deployed my own stack on Hetzner with nginx, TLS and SSH-key access). For the trial: install OpenClaw pinned, local npm, dedicated non-sudo user, under systemd; harden with SSH key-only, UFW, fail2ban, everything loopback-bound except 22; monitoring with CrowdSec + Netdata; secrets in env at 0600, never in unit files or shell history. Hand back the working box, a short runbook, and a Loom of the key parts. Context project: GitHub - PANTOK-cmyk/bean: Telegram Business bot that auto-replies on your behalf with an LLM + RAG (Postgres/FAISS). aiogram · OpenAI/Anthropic. · GitHub , a self-hosted LLM bot on Postgres. EU timezone, same-day reachable. Trial quote: $250 fixed, then $40/hr for ongoing work.

16

Hi James - for the Telegram photo flow, I would have Telegram pass file_id, sender, timestamp and any job/order reference into n8n, fetch the file through the Telegram Bot API, store it in durable storage with a path like job-id/date/message-id, then validate the job in monday.com and write the image URL plus Telegram message ID back to the matched item. Ambiguous matches should go to a review queue, not be guessed. Relevant link: I do not have a public OpenClaw case study to honestly share yet; I would prove fit on the trial with the hardened box, runbook and screen recording. Fixed trial quote: USD 149. Ongoing: USD 45/hr. Timezone: UTC+8; same-day urgent acknowledgement, often 1-2 hours if I am at the keyboard.

17

Hi James

  1. Getting the photo stored and linked to the right job.
    First thing, I’d pull the photo off Telegram straight away (download it — file_ids expire), drop it into object storage (a self-hosted MinIO bucket fits your setup) under a UUID, and log the raw inbound — sender, timestamp, file hash — before anything else, so a photo is never lost even if matching fails. For the link itself I treat the model as a suggester, not the source of truth: I pull that sender’s open jobs from the DB and match deterministically first — a job number in the caption wins; if there isn’t one, I OCR the text printed on the sign and fuzzy-match it against that real candidate list. I’ve actually shipped this exact “snap a photo → recognise it → turn it into a structured record” flow in production — I have an AI expense-tracker on the App Store (FernBudget) where you photograph a receipt and it visually extracts it straight into a structured entry ( ‎FernBudget: Expense Tracker App - App Store ); swap “receipt” for “sign” and “bookkeeping entry” for “job link” and that’s your problem. The agent can only ever pick a real job ID — it can’t invent one — and the actual write goes through a validated, idempotent tool that checks the job exists and is still awaiting a photo. A single high-confidence match auto-links (bot replies “linked to JOB-1234 ✓, reply NO if wrong”); anything ambiguous, the bot shows the sender their open jobs as buttons to confirm — never a silent guess. The photo URL plus how it matched (method + confidence + auto/human) is written onto the job record, and anything unmatched goes to an “unassigned” queue rather than being dropped.
  2. Closest things I’ve built.
    • An open-source OpenClaw security/hardening audit tool I wrote: GitHub - Pelican0126/-openaclaw-security-check: OpenClaw 安全巡检脚本：Gate/Inventory/APIKey 泄露审计，默认脱敏与问题摘要，支持隔离/备份删除。 OpenClaw security audit scripts: Gate/Inventory/API key leak scan, masked reports & concise findings, quarantine/backup/delete support. · GitHub — one-shot checks over an OpenClaw box (gate + inventory, API-key leak scanning, masked reports, quarantine/backup). That’s the hardening + secrets-hygiene side of exactly this job.
    • Vision / on-device OCR: GlyphScan, my open-source on-device OCR + fuzzy-matching library ( GitHub - Pelican0126/GlyphScan: On-device, OCR-aware fuzzy matching that finds the one record your camera is pointed at — pure Swift, CJK-first. · GitHub ), plus the App Store expense-tracker above (photo → visual recognition → structured data) — together they cover the whole “photo → extract → link to a record” chain.
  3. Pricing.
    Trial: US$380 fixed. Ongoing: US$40/hr.
  4. Timezone / response.
    Taipei (UTC+8) — only a couple of hours behind you, so good daytime overlap. For urgent breakages I can usually be on it within an hour or two during the day, and I keep alerts on after hours. Screen-recording the work is no problem — I work that way anyway.
18

Hey @James_Nation

I’ll harden a fresh Hetzner VPS into a production-ready OpenClaw host with proper isolation, monitoring, and documentation you can reference and repeat.

Here’s the approach:

• OpenClaw under systemd, pinned version, dedicated non-sudo user, local npm
• SSH key-only, UFW, fail2ban, loopback-bound services except:22
• CrowdSec + Netdata for monitoring (open to discussing alternatives with reasoning)
• Env-file secrets at 0600, never in unit files or history
• Runbook + Loom walkthrough of every significant step

On your Telegram photo question: the agent receives the photo via webhook, extracts the job reference from the caption or replied-to message, uploads the image to object storage (S3/R2) with a job-keyed filename, then writes the URL plus metadata back to the monday.com item via API — so the photo lives next to the job record, not buried in chat.

Based on your requirements, here’s my relevant work:

Let’s schedule a call to discuss the details further.

19

Hi James,

  1. I’d have the Telegram agent capture the file_id, sender/chat, timestamp, and any job code/caption, then let n8n download the image via Telegram getFile and store it in object storage with a deterministic path like jobs/{job_id}/finished/{message_id}. For matching, I’d first prefer a QR/job number in the caption or sign label; if missing, I’d use OCR/vision extraction and cross-check against active jobs in monday.com or the production DB. Only after a unique match would the workflow update the job record with the image URL, Telegram metadata, and an audit log; ambiguous cases go to a small human review queue with Telegram buttons instead of guessing. That keeps the AI useful, but not authoritative.
  2. Closest relevant link: available on request.
  3. Fixed trial quote: USD 220 fixed for the VPS hardening + OpenClaw install + short runbook + screen recording. Ongoing rate: USD 40/hr.
  4. Timezone: Asia/Kuala_Lumpur UTC+8. For urgent fixes, usually same day; for pre-agreed on-call windows, I can usually jump in within 2-4 hours.

I work best on reliability-focused automation: repeatable commands, clear security boundaries, secrets kept out of unit files/history, and handoffs another person can re-run. I’m comfortable doing this as a fixed-scope paid trial and documenting the significant steps with notes + recording.

20

Hey, refreshing to see someone who’s actually built and running a stack before hiring — makes this a lot easier to scope.

To your four points

1. Telegram photo to job record
When the photo comes in via Telegram webhook in n8n I’d extract the file ID, pull the binary using the Telegram node, then push it straight to your storage layer (S3, Cloudflare R2, whatever you’re using) and write the public URL back to the monday.com job record using the job identifier pulled from either the caption or a preceding message in that conversation thread. If staff aren’t consistent with captions I’d add a quick confirmation step that pings them back asking to confirm the job number before it files anything. Keeps the data clean without adding much friction.

2. Most relevant thing I’ve built
The ClickHouse node — DevOps adjacent, database operations, self-hosted environment. That’s the closest match to the infrastructure work you’re describing. basitq.cloud/nodes

3. Quote
I’d want to see the VPS and scope the hardening task properly before giving you a number that’s actually honest rather than just competitive. Happy to quote after a 15 min call. Ongoing hourly we can discuss privately based on complexity.

4. Timezone and response time
I work US timezones. For urgent stuff I’m usually responsive within a couple hours during working hours, same day outside of them.

Book a call here and we can get into it: https://calendar.app.google/qDYxm4uA6cpScfTD7