Systems in Production

From Workflow to Working System

Each case began as a manual process in a real business: the problem, the system we built, and the result.

Selected client work

Systems built around real business problems

Each project started as a bottleneck inside a real business. Here is what we built and what changed.

EdTech / Career Development

Outcome

Career guidance + AI learning system

Problem: Students lacked clear career direction and a way to turn academic effort into job-ready outcomes.
Solution: An AI-powered career intelligence platform that maps skills, personalizes learning paths, and guides decisions based on individual goals.
Result: Built a product that turns fragmented career guidance into a structured, intelligent decision engine.

Public Sector / GovTech

Outcome

Public service deployment

Problem: Public employment services needed more intelligent, scalable support for users navigating a complex ecosystem.
Solution: AI-assisted service design and deployment for a national career and employment support experience.
Result: Extended digital support into a trust-heavy public environment with better access and improved information flow.

Startup Ecosystem / Platform Engineering

Outcome

Marketplace module shipped on a new modular platform

Problem: Bizowl's product had outgrown its original architecture. New features had become hard to add, and users were left unsure what the platform was for and what to do next.
Solution: A new modular platform foundation with reusable core services, a marketplace module, a redesigned and more intuitive user experience, and a careful migration of existing client and partner data.
Result: Bizowl now has a marketplace module shipped on a more structured platform, with its existing business data carried over and a clearer experience for users.

Product Engineering / Custom Platform Development

Outcome

Custom affiliate management system live in production

Problem: Ace10x needed to manage partners, track referrals, and calculate commissions, but off-the-shelf affiliate platforms came with broader feature sets, integrations, and recurring costs of $50-$100 a month than its use case required.
Solution: A focused affiliate management module built natively into Ace10x's existing product on Django, PostgreSQL, and Redis, covering tracking, a commission ledger, partner management, payouts, reporting, and fraud prevention.
Result: A custom affiliate system live in production, with reported 99% tracking accuracy and no separate third-party affiliate platform needed for core workflows.

Specialised Services / Lead Intelligence

Outcome

About ₹0.80 estimated cost per researched lead

Problem: Two employees spent their time manually researching incoming leads, roughly ₹40,000 a month in staffing, and the workload grew with every new lead.
Solution: A purpose-built research pipeline: Serper for web search, Jina AI for content extraction, and a lighter Gemini model for classification, run by two workers on a Django and MongoDB backend.
Result: Estimated research cost fell to about ₹0.80 per lead, from roughly ₹2-₹3 with a web-search-agent approach and ₹5-₹8 with general-purpose AI research.