I build AI agents and automations for the messy work people keep postponing. Most projects start with a spreadsheet, an inbox, or a workflow nobody wants to touch, and end as software the team can actually run.
How I help
You do not need a pitch deck. A rough description of the workflow is enough to start.
Map the current steps, risks, tools, and handoffs before deciding what should be automated at all.
Build the first useful version with logs, review points, and a clear path for the edge cases.
Create the screens people need to review, correct, and run the automation day to day.
First step
We map the steps, find the automation points, define the human review rules, and decide honestly whether an agent is worth building.
You bring
You get
Selected work
Each one shows the workflow, the human review points, the tools involved, and the guardrails worth inspecting before trusting automation.
Researches inbound leads, drafts the first reply, and prepares the CRM update so sales never starts from a blank screen.
Turns rough briefs into drafts, checks coverage, and readies articles for a human editor to review.
Reads new tickets, classifies urgency, drafts replies from the knowledge base, and escalates edge cases with full context.
Internal search that answers from real documents and shows the passage each answer came from.
Reads invoices, pulls out the boring fields, catches odd numbers, and leaves finance only the parts worth checking.
Reviews resumes against the actual role, explains the tradeoffs, and hands recruiters a cleaner shortlist to argue with.
Follows up with warm leads, answers the FAQs, updates CRM stages, and pings sales the moment a buyer is actually ready.
Pulls decisions, risks, and follow-ups out of call transcripts before everyone forgets what was agreed.
Profile site for a legal practice in Faridabad. Design, build, and content structure.
Nothing in that category yet. Try another filter.
Expertise
The work usually crosses a few lines: product thinking, backend plumbing, prompts, interfaces, and the unglamorous parts that keep things running.
Agents for jobs that already exist: research, support, lead scoring, messy handoffs, and the parts that need judgment before action.
FeaturedAutomations that connect the tools you actually use: CRMs, inboxes, databases, docs, queues, and the occasional stubborn API.
FeaturedOpenAI, Claude, local models, retrieval, vector search, and prompts written around the job itself.
FeaturedLaravel, Node, APIs, queues, databases, auth, billing, dashboards. This is the layer an AI feature usually breaks on first.
Screens for reviewing, correcting, and approving AI output. Most agent projects need more product design than they need chat.
Deciding what is worth automating, what needs a human in the loop, and what should probably just be left alone.
My story
I started building for the web while studying Computer Science at Delhi University. I left the degree, kept the work, and never really stopped learning by shipping. These days it’s mostly AI agents, but the old rules still hold: clear data, sane interfaces, good failure paths, and no magic where a boring check would do.
Clients usually come to me when they need one person who can shape the workflow, build the agent, design the review screen, and wire the whole thing into the tools they already use.
AI Agents & Automation · Remote
Building agents and automations for founders and small teams. Usually OpenAI or Claude, often n8n, sometimes LangChain or RAG, almost always tied back to a Laravel or Node product.
Full Stack Web Developer · Faridabad, India
Built and maintained the platform while also teaching online. That combination taught me a lot about software that has to work for real people, not just developers.
Full Stack Developer & Startup Consultant · Baltimore, MD
Joined as a full-stack developer and grew into a consulting role on a C2C marketplace serving students and tutors across the US.
Head of IT Department · Faridabad, India
Started as a developer and later ran the IT side. Small team, lots of responsibility, and very little room for overengineering.
Independent client work · Remote
Left Delhi University's CS program in 2013 to learn by shipping. Hundreds of client conversations later, still learning every day.
Education
2013 - Present
Web development, design, product thinking, and lately a lot of agent workflows. Mostly learned by building things and fixing what broke.
2011 - 2013
Bachelor of Computer Science · Delhi, India.
2000 - 2011
High School · Faridabad, India.
Let’s build with AI
I’m taking on a few AI agent and automation builds for 2026. If there’s a task your team keeps doing by hand, I can help work out whether it should become software or stay with a person.