A gathering place for clever ideas, thoughtful products, and open-source tools built for the everyday lives of real people — by Arindam & Sushmita.
Adda is the Bengali tradition of friends gathered over tea — talking, arguing, dreaming, making sense of the world together. It is unhurried and irreverent. It builds nothing on a roadmap and somehow builds everything that matters.
Yukti is the Sanskrit word for the clever turn of thought — the elegant idea that solves what force cannot. The workaround. The unexpected angle. The recipe that just works.
This place is built for both. For products designed with care. For applied AI that earns its keep. For multi-agent systems that route, retrieve, and reason. For code shared openly so the next builder doesn't start from zero. For tools that help an everyday person — a small-shop owner, a teacher, a parent, a community worker — do what they already do, a little easier.
It is a workshop, a notebook, and a chai-time conversation — all at the same address. We build because good tools belong to everyone. Pull up a chair.
Different disciplines, shared philosophy. Products that respect the people inside them, and applied AI that earns its place by being useful in the real world.
Building enterprise software for nine-plus years across Zoho · Qntrl and Ericsson, designing workflow orchestration platforms for Fortune-500 enterprises and large-scale operations. Currently pursuing a Doctorate in Business Administration, researching how AI and process management reshape organizational decision-making — not just optimize it. Published IEEE author across three papers on IoT and smart infrastructure. The belief that drives the work: the best AI strategy is the one that makes complex systems feel simple for the people inside them.
Applied AI Engineer with eight-plus years building scalable backend systems and two-plus years deep in production-grade GenAI applications. Currently IT Analyst at Tata Consultancy Services, engineering payment microservices and integrating ML model inference into real-time decision workflows. Specializes in multi-agent AI systems, Retrieval-Augmented Generation, and LLM-based solutions using LangChain and LangGraph. Currently pursuing a Post Graduate Program in AI/ML at The University of Texas at Austin. The throughline: AI systems that are explainable, reliable, and deployment-ready — for people who actually need them.
A team collaboration platform built on one insight: the problem isn't missing data. The problem is fragmented data.
Your team's intelligence lives in fragments too. The decision rationale is in a chat thread from three weeks ago. The task it created is in Projects. The document referenced in the thread is somewhere in Notes. The customer email that started it all is in someone's inbox. We have all the pieces. We just can't connect them.
Nine years of building enterprise SaaS taught me this is the real reason teams feel slow. Not missing tools. Not bad processes. Just fragmentation. So Synergy was built on one principle: every message, task, and decision shares the same underlying data model. A message knows which task it spawned. A task links to the decision behind it.
AI synthesizes across all of it — instead of humans copy-pasting summaries between tabs. Gemini summarizing a channel in two seconds isn't impressive because it writes well. It's impressive because no human had to stitch the context together first. AI is most valuable as connective tissue, not creative output.
The hardest part of building wasn't the code. It was resisting the urge to ship features and instead obsessing over the insight underneath. Teams don't need another tool. They need their existing tools to stop being silos. Solve the underlying problem, not the surface complaint.
Two disciplines sharing one address — product architecture and applied AI engineering — connected by a single philosophy: build systems that respect the human inside them and give back to the community outside them.
Enterprise software is famously complicated. The work is finding the rare angle where complexity moves into the machine and simplicity moves into the hands of the people using it. Orchestration, hybrid systems, low-code builders that empower non-technical teams.
Production-grade GenAI: RAG pipelines over real document corpora, LangGraph-orchestrated agents that pick the right tool for the right question, vector search that returns grounded answers with source attribution. Built for explainability and reliable deployment.
The best AI in enterprise isn't the loudest. It's the layer that quietly removes drudgery, leaves judgment where it belongs, and earns trust by being legible.
Payment microservices, event streaming, ML inference APIs in production — backend engineering that treats reliability and observability as features, not afterthoughts.
A home GPU lab. Self-hosted models. Multi-agent prototypes. Tools that exist because curiosity is its own reason — and because someone else might need them too.
Doctoral research on human-centric AI. Three IEEE papers published. PGP-AI at UT Austin. Field notes from product and engineering trenches.
Bengali cooking, weekend hardware tinkering, half-finished essays — the side practices that quietly teach the main ones.
Arindam designs how teams of humans coordinate across enterprise software. Sushmita designs how teams of AI agents coordinate to answer hard questions. The work looks different on the surface — but the underlying craft is identical: systems should reduce fragmentation, not add to it, and what gets learned along the way belongs in the open, where the next builder can use it.
The most useful technology of the last twenty years has come from people who built openly and gave away the recipe. Linux. Python. PostgreSQL. The internet itself. We grew up on those gifts, and the only fair response is to keep passing them forward.
Yukti Adda is built on a simple commitment: the tools we make are tools we give back. Not as marketing. Not as charity. As a working principle — because what gets learned in one corner of the world should help the next person, in the next corner, who is trying to solve a slightly different version of the same problem.
We're especially interested in tools that touch everyday life: small-business owners reconciling messy invoices, teachers handling paperwork, parents managing family logistics, community workers tracking who-needs-what. The unglamorous middle layer of life that big AI labs don't optimize for. That's where the next wave of useful AI lives.
We build for India because it's home, and we build in the open because every line of code we write should be readable, forkable, and improvable by anyone — anywhere — who wants to take it further.
Tools, prototypes, documentation, and field notes published openly unless there's a clear reason not to.
Not just developers, not just enterprises. Real tools for real lives — shopkeepers, teachers, parents, community workers.
Designed for Indian realities — multilingual, low-bandwidth-friendly, affordable — and available to anyone, anywhere.
Not the grand gesture. The quiet utility, finished and given away — then the next one. And the one after that.
If you're building enterprise AI, multi-agent systems, intelligent workflows, or open-source tools that help everyday people live a little better — the adda has room for one more.