Archit Agarwal

I build production AI systems that turn unstructured business signals into operational intelligence. Much of the information that runs a business never reaches a system of record. It lives in voice instructions, video and everyday conversation. My work is about capturing that layer and making it structured enough to act on.

That idea led me to found Corterex, where I designed and built Cerlut, an operational intelligence platform for small and medium businesses. Owners delegated work through spoken instructions that left no record of what was done or what had slipped. Cerlut accepted multilingual voice input, extracted structured tasks, resolved each one to the right team member and tracked it to completion. The hardest design question was trust. One wrong task could cost an owner’s confidence, so the model acted as a drafting layer, with a human-in-the-loop approval step before anything reached the team.

Around 7 businesses used Cerlut at MVP stage, and I presented it to 15 to 20 owners. The key finding was behavioural. Owners resisted changing habits they had relied on for years, and I parked the product on that insight. It taught me that adoption is a design problem as much as a technical one.

I applied the same thinking to inbound demand for a hardware merchant, as paid freelance work. Orders, enquiries and complaints from a chatbot and phone calls were parsed by an AI agent into structured records on a central dashboard. The manager moved each order from receipt to delivery, while the owner had a single view of every lead and order.

I also designed a multi-agent creative intelligence system for D2C brands and agencies. It performs multimodal analysis of competitor video ads, with specialised agents assessing visual style, narrative structure, audio and engagement in parallel. A synthesis layer turns their findings into a strategic report, ad scripts and prompts for new creative. I demonstrated it to marketing agencies in Australia and India.

My foundation in production software came from 1.5 years at CRED, in frontend quality engineering for CRED Pay, the company’s second biggest revenue line. I owned release quality for its new BBPS (Bharat Bill Payment System) feature, where a defect reaches users as a payment problem. I joined as an intern and was promoted to consultant within 4 months.

I hold a master’s in IT with a specialisation in AI. In Melbourne, I led a team of six during an internship at Regeneration Projects, delivering Two Bays, a pilot application for university students.

I’m looking for remote roles in applied AI and LLM engineering, and ML engineering roles centred on LLM systems. I do my best work turning complex, unstructured processes into dependable AI workflows, with clear boundaries between what the model decides and what stays with people. Longer term, I want to work closer to the model layer, and I study model architecture and training alongside my work.