What RAG actually needs to work in production
Retrieval quality, evaluation, and product context matter more than picking the most impressive model.
Software fordifficult problems.
We design and build software, products, and intelligent systems for teams who need more than implementation — they need someone who understands the problem well enough to own it.
projects shipped09
products in motion02
senior team01
INBuilt in IndiaTrusted worldwide
Engineering, not just implementation.
We work directly with clients to understand the problem, design the system, and build software that lasts beyond the first release.
No handoffs between the people who talk to you and the people who build it — one team owns both, from the first architecture decision to what happens after launch.
What we build
Organized around the problem, not the tech stack.
AI & intelligent systems
We design retrieval systems, copilots, intelligent workflows, and AI products around the information and decisions your team already has.
LLMs · RAG · Evaluation · AgentsProduct design & engineering
We combine product thinking, UI/UX design, and full-stack engineering so the interface and the underlying system are designed as one product.
UI/UX · React · TypeScript · Node.jsBackend & systems
APIs, data models, and services designed around real domain boundaries, predictable operations, and the team that will own them after handoff.
Go · Node.js · PostgreSQL · RedisAutomation & workflows
We map the process, connect the systems involved, and build workflows that are observable, retryable, and easy to trust.
Python · APIs · Workflows · IntegrationsCloud, infrastructure & security
We design deployment infrastructure and security practices that keep systems reliable, protected, and understandable as they grow.
Docker · CI/CD · Cloud · SecurityModernization & architecture
Incremental architectural improvements that reduce risk while preserving the business rules and hard-won knowledge inside existing systems.
Architecture · Migration · RefactoringSelected work
Systems shaped around the people using them.
Case files · selected
Case fileProduct · AI
A search layer for the company's scattered memory
Anvaya is an AI-powered enterprise assistant that provides natural-language access to organisational data, knowledge, and connected tools. It can handle multi-step analytical queries, retrieve information across multiple sources, execute authorised operations, and adapt its capabilities to each user's role and permissions. Built as a multi-tenant PWA with configurable AI providers, caching, and controlled data retention, Anvaya turns complex enterprise information workflows into a single conversational interface.
Open case ↗Case fileProduct · EngineeringMaking government schemes easier to discover, understand, and access
Yojana Sahayak is a multilingual platform that makes government welfare schemes easier to discover, understand, and access. We built the full-stack platform, an automated government-scheme data collection pipeline, and an AI assistant supporting text and voice interactions across eight languages.
Open case ↗Case fileSystems · BackendFast, reliable notification delivery across every channel
Sandesh is a lightweight multi-channel notification delivery system built to reliably send push notifications, email, and SMS while supporting priority-based queuing, retries, scheduling, templating, and failure handling.
Open case ↗Case fileAutomationComputer vision for safer workplaces
Sentinel is a real-time computer-vision system for automated PPE compliance monitoring. It detects workers, helmets, and safety vests from images and camera feeds and identifies missing protective equipment as potential safety violations.
Open case ↗Products in motion
Things we build because we want to use them.
How we work
Clarity before velocity. Momentum after.
01
Understand
The business and technical problem, before any tooling decisions.
02
Model
Users, data, constraints, and how the system needs to behave.
03
Design
The simplest architecture that actually solves the problem.
04
Build
Implement iteratively, in reviewable, shippable pieces.
05
Measure
Validate against real outcomes, not assumptions.
06
Improve
Keep improving the system after it ships.
Engineering notes
Writing about what we build.
Choosing boring technology for a system that has to last
Why proven tools are often the most ambitious choice when a team has to own the system after launch.
Modernizing a system without a full rewrite
A practical approach to replacing the parts of a legacy system that are actually causing pain.
Building something
difficult?
Tell us what you're building, improving, or trying to figure out.