Client, product & AI work
Projects led by Arpit Desai
Client programmes I’ve led as Architect, Program Manager, Technical Lead and Senior Engineer—plus AI solutions I’ve built: Copilot Studio bots, connector frameworks for the SDLC, and multi‑agent systems for engineering delivery and software testing.
Enterprise reward points platform
Working with a 60+ person team for a US‑based client on a large‑scale rewards program, owning the front‑end architecture and delivery and providing solutions for complex UI flows.
Digital platform for manufacturing operations
Canada‑based product for the manufacturing industry, built with Vue.js and Laravel. I worked as a Senior Developer handling both front‑end and back‑end, as well as direct client communication.
Appraisal & banking platform
Large US‑based platform in the appraisal and banking domain, serving thousands of requests per second and millions of users. I handled multiple integrations and daily coordination with clients.
Core banking applications
Served as program manager for a team of 20+ developers building core banking applications for a large US financial institution. I worked from the United States for more than three years, aligning delivery with stakeholders while contributing as delivery lead and architect on critical parts of the programme.
Online booking & tracking systems
Delivered projects such as an online hotel booking system, a labour tracking system and multiple chat and lab management applications using Node.js, PHP, MySQL and MongoDB.
E‑commerce & communication tools
Built custom e‑commerce products and real‑time communication tools early in my career, strengthening my foundations in full‑stack web development, UI/UX and performance.
AI · Copilot Studio · Agentic systems
AI solutions I’ve delivered
Selected work across conversational AI, enterprise connectors for LLMs, and autonomous multi‑agent systems for engineering and quality practices.
HR chatbot for employees
Created an HR‑focused chatbot in Microsoft Teams using Copilot Studio, so employees can ask questions and complete common HR tasks through a governed conversational experience instead of hunting through static portals alone.
LLM‑ready connector framework
Prepared a connector‑based framework for enterprise use so LLMs can pull trustworthy context from Git, Swagger, Datadog, Jira, and Azure services. The aim is to support day‑to‑day SDLC work— triage, design discussions, reviews, and operational insight—with grounded, tool‑aware assistance rather than one‑off prompts.
Engineering team agentic AI system
Designed an autonomous agentic system that mirrors a software organisation: agents for technical product owner (TPO), team lead, developer, QA, and DevOps. From a single brief, the agents collaborate to drive end‑to‑end delivery—from clarification and implementation through validation and release‑oriented activities—under explicit guardrails you define.
STLC agentic testing platform
Built a knowledge‑ and context‑graph–based STLC solution where different AI agents help create, maintain, and fix test cases as needs and workflows evolve. The system ties testing artefacts to product context so suites stay aligned with changing requirements and execution feedback instead of drifting after each release.
How I think about projects
Principles I bring to my work
These principles shape how I design and deliver software systems with teams, regardless of the specific tech stack.
Ship small, learn fast. I aim for thin‑slice versions of ambitious ideas – something real that we can put in front of users quickly to learn from their behaviour.
Evidence over intuition. Intuition is a great starting point, but I try to anchor decisions in data: evaluation sets, experiments and clear qualitative feedback from the people who will rely on the system.
Responsible by design. I care about failure modes and edge cases early. That means talking about guardrails, escalation paths and what “safe enough” really means for your context.