Posts
AI/ML Engineer building production LLM applications and agentic AI systems
Most of my time outside work right now goes into Aaptam, a product I am building properly, from research to release. It is still in development, so this is a first look rather than a launch. Aaptam is an expense tracker built around one idea: logging money should take seconds, not discipline. Snap a photo of a bill and it reads the merchant, every line item and the tax, ready for you to review. The part I am most excited about is Tam, the AI assistant at its centre. Tam is not a chatbot that writes paragraphs of advice. It has 35+ tools it can use to act on your records, such as adding transactions, adjusting budgets, tracking goals and checking what bills are coming up. Its answers come back as interactive widgets, like a budget progress card or a list of upcoming subscriptions, instead of walls of generated text. Anything destructive needs your confirmation, and every feature also works by hand for people who would rather not use AI. A few principles are shaping it: Free to track, with no ads. The app should never have a reason to sell what it learns about you. Your month starts on payday. Budgets and reports follow your salary date, not the calendar. UPI, cards and cash in one place, with transfers between your own accounts never counted as spending. Local-first. Records start on your device, with optional backup to your own Google Drive. Under the hood it is a React application shipped to Android with Capacitor, backed by Supabase, Firebase and serverless functions. Aaptam is a standalone product with its own brand, built independently of everything else I have made. The Android app is in closed testing, and I will share a proper deep dive when it launches.
AI/ML Engineer building production LLM applications and agentic AI systems
I have open-sourced Pludo Coder, the résumé-to-portfolio builder I launched in August 2025. The full codebase is now public under the AGPL-3.0 licence: github.com/pludoai/Open-Pludo-coder↗. Before publishing, I cleaned up the repository: secrets moved out into configuration, an example environment file added, and a README that explains how to run it yourself. The reason is focus. My time now goes into Aaptam, a product I am building properly for production, and Pludo Coder deserves better than sitting idle. If you want a starting point for AI-generated portfolios, with a visual editor and a chat-based editor already in place, fork it and take it further.
Pludo Coder
AI/ML Engineer building production LLM applications and agentic AI systems
Honoured to receive the Champion Award at the Aalee Awards 2026, the highest individual recognition at Aalee. The award slide summed it up in one line: "The one you call when it has to be done." The trophy puts it more formally: in recognition of delivering when it matters most, taking ownership in critical moments and setting the standard through action. Since joining Aalee in November 2025, most of my work has gone into the core systems of Aalee Assist, our AI support platform. This award is for how that work got delivered when the pressure was on, and it is the recognition I value most so far. Thank you to the leadership, my mentors and the whole Aalee team for trusting me with responsibility bigger than my comfort zone. Still learning, and still building, every day.
AI/ML Engineer building production LLM applications and agentic AI systems
Earlier the same morning, I received the Rising Star Award, presented by Pheonix Solutions, Aalee's parent company, for my work with the Aalee team. The plaque reads: "Awarded for emerging excellence and a strong trajectory in automation and reliability." Automation and reliability are the parts of the job that rarely show up in a demo: tool-calling workflows that behave the same way every time, and escalations that reach a human exactly when they should. It means a lot to see that side of the work recognised. Thank you to the leadership at Pheonix Solutions and Aalee for the recognition.
AI/ML Engineer building production LLM applications and agentic AI systems
Aalee Assist is the product I spend my days on at Aalee. It is a no-code platform for creating, training and deploying AI assistants that answer from a company's own knowledge, and it is where most of my engineering work goes. The problem it solves is a familiar one. Customer conversations arrive on a website chat, WhatsApp, Telegram, Instagram, Messenger, Slack, email and Mattermost, each with its own inbox and its own history. Bolt-on chatbots answer what they recognise and dead-end the rest, so the handover to a person comes late and without context. Assist puts all of those channels behind one assistant and one shared dashboard: One assistant, eight channels. The website widget plus seven messaging channels, all in a single thread view. Escalation decided by intent. Through tool calling, the assistant decides when to answer, when to hand over, and which agent should take the conversation, with the full history carried across. Bring your own key. Teams connect their own accounts across 15+ LLM providers and switch models without a migration. Grounded in your knowledge. Assistants are trained on the company's own content rather than answering from general knowledge. I lead Assist and built most of it, including the LLM and tool-calling layer, provider routing, the embeddable web widget and the agent dashboard. The stack is Python and FastAPI on the backend, React and TypeScript on the front, with Vertex AI, Redis and WebSockets underneath. You can try it at assist.aalee.ai↗.
Aalee Assist
AI/ML Engineer building production LLM applications and agentic AI systems
I was invited to lead a hands-on workshop for more than 80 final-year students of the Department of Computer Applications at MIT College of Arts and Science for Women, Musiri. Instead of theory, we worked through how real web applications get built and shipped. We covered how a side project grows into a SaaS product, how to plan a final-year project so it is worth showing to an employer, and what to expect when starting a career in tech. The best part was the questions. Students who are about to graduate want to know what the work actually looks like, and answering that honestly is the most useful thing a session like this can do. Thank you to the Department of Computer Applications for the invitation and a warm welcome. Explaining your work clearly to people just starting out is a skill in itself, and I learned as much from the room as I shared.
AI/ML Engineer building production LLM applications and agentic AI systems
Pludo Coder turns a résumé into a live portfolio website. It reads the document, pulls out experience, projects and skills, and generates a responsive site you can publish in one click. After that, there are two ways to keep editing: a visual editor for text and styles, or an AI editor where you simply describe the change you want. Every portfolio comes with its own themes, so the same content can look completely different in a couple of clicks. It went live in August 2025 as part of Pludo AI, my independent product studio.
Pludo Coder
AI/ML Engineer building production LLM applications and agentic AI systems
In July 2025 I opened a private beta of Pludo Builder, a no-code builder for AI customer-support agents. You describe the agent (its name, brand and role), choose an AI provider and model (OpenRouter, Google Gemini or Z.AI), and get a chat widget you can embed on any website. Every agent also gets its own hosted page on a subdomain, so a business without a website can still use it. Access was invite-only through a waitlist, which kept the beta small enough to learn from every conversation. The product has since been discontinued, but building it end to end, from model routing to the embeddable widget, is where I learned what it takes to put an AI agent in front of real users.
Pludo Builder
AI/ML Engineer building production LLM applications and agentic AI systems
Placed 2nd in the Frontend Development Hackathon at Karpagam College of Engineering, Coimbatore. It was an on-the-spot build, judged live on March 15, 2024: my team (M. Pravin, S. Manoj and I) designed and built a website from a blank screen in one sitting.
