Workflow Automation
- 49 workflows built, ten live on daily, scheduled and real-time triggers
- Full LMS student lifecycle automated across 13 branches
- Zero manual kick-off — every run is scheduled or event-driven
Building production automation, retrieval-grounded agents, and the internal platforms they feed.
I build automation that runs unattended against real operational data — student lifecycles, fee collection, telephony reporting — and the internal platforms that make its output usable by the people doing the work.
I started as a React JS Developer at SNVA Ventures in September 2024, building client-facing applications and internal dashboards. I joined SkillCircle in February 2025 as an IT Executive, running hosting for 1,200+ domains across cPanel, DNS, SSL and Linux. I began automating the repetitive parts of that job; the automations became core operations, and in January 2026 the company created the AI & Automation Engineer role for it.
Ten of 49 workflows running on daily, scheduled and real-time triggers across Edmingle, Google Workspace, MCube telephony and WHMCS — no sandbox, no manual kick-off.
Retrieval-grounded answers with structured output and an explicit escalation flag, so refunds, complaints and account issues route to a human instead of receiving an invented reply.
Next.js and NestJS applications on Supabase — job portal, admission dashboard, sales portal — owned from data model and migrations through interface and deployment.
Happy to walk through any workflow or platform on this site, live.
Technologies and tools I use to build production systems.
49 workflows built, eight live on scheduled, webhook and real-time triggers.
Qdrant vector store, Gemini embeddings, 800/80 chunking, top-8 retrieval.
Structured output with explicit human-escalation branches.
Ad-creative generation and grounded support responses.
Retrieval chain and output parsing inside the support agent.
Fee reminders and outbound messaging before the n8n migration.
Focused on production automation, grounded agents and the systems they integrate with.
Shivam Gupta is an AI & Automation Engineer based in Noida, India. He builds production workflow automation, retrieval-grounded AI agents, and the full-stack platforms those systems feed. He started as a React JS Developer at SNVA Ventures in September 2024, then joined SkillCircle in February 2025 as an IT Executive managing hosting for 1,200+ domains, and moved into the AI & Automation Engineer role in January 2026 after the automations he built became core operations.
Shivam Gupta has built 49 workflows at SkillCircle, ten of them running in production — eight in n8n, the rest on other tooling. They cover the LMS student lifecycle across 13 branches — daily onboarding provisioning and scheduled archival — plus fee-reminder emails, real-time certificate generation and delivery, Canva seat approvals, and a nightly sales-performance alert that joins counsellor rosters, walk-in logs and live MCube telephony data. Every one runs on a schedule or a webhook with no manual start.
The agent answers inbound support queries from SkillCircle's brochure corpus. The documents are chunked at 800 tokens with 80 tokens of overlap into a Qdrant vector store using Gemini embeddings, and the top eight passages are retrieved per query. The model is constrained to answer only from those retrieved passages and returns a structured object containing both the reply and an escalation flag, so refunds, complaints and account-specific questions route to a human instead of receiving an invented answer.
Shivam Gupta works primarily in n8n for workflow automation, with Qdrant, Gemini and LangChain for retrieval-grounded agents. On the application side he uses TypeScript with Next.js and NestJS, backed by Supabase and PostgreSQL, with Redis and BullMQ for queued work. He also runs the infrastructure underneath: Linux hosting, cPanel and WHM, DNS, SSL and Cloudflare across more than 1,200 domains.
Shivam Gupta builds automation in three stages. First he traces the manual process end to end with whoever performs it today, including the steps nobody documented. Then he builds a read-only version, using diagnostic probes to confirm the data is shaped as expected while write operations stay disabled. Only once inputs are verified does he enable writes against a small slice, check the output by hand, then widen. Finished systems run unattended and report only exceptions.
Yes. Shivam Gupta is open to AI Engineer and AI & Automation Engineer roles, remote or in Delhi NCR. He can be contacted through the form at https://shivamg.in/contact, and his work is documented at https://shivamg.in/projects with architecture write-ups at https://shivamg.in/case-studies.
Let’s talk about the manual work you’d rather not be doing next quarter.
Let’s Collaborate