Full-stack development
Design and build web applications, APIs, backend services, and internal tools with modern TypeScript-based stacks.
Kashmere Labs Pte. Ltd. · Product and AI engineering
From full-stack applications and agent integrations to data labelling and evaluation, Kashmere Labs takes scoped requirements through implementation and deployment.
01 / Services
Plan, build, integrate, and deploy the software, agent, and data layers required by the product.
Design and build web applications, APIs, backend services, and internal tools with modern TypeScript-based stacks.
Build A2A gateways, tool-using agents, assistant workflows, and retrieval integrations that connect with existing products and APIs.
Create annotation workflows, review queues, quality checks, and evaluation datasets for training and validating AI systems.
Implement image, video, audio, and media-automation workflows using composable models and provider APIs.
Build embedding pipelines, semantic search, related-item recommendations, and retrieval-backed knowledge experiences.
Package services with containers and deploy application stacks with runtime configuration, logging, monitoring, and release automation.
02 / Delivery system
A high-level delivery pattern used across custom engagements and the products operated by Kashmere Labs.
03 / Products
Live applications and open-source building blocks built around the Samsar generative media platform.
Flagship platform
samsar.one
Samsar creates videos up to three minutes from prompts or image lists, then supports scene rerolls, media editing, semantic search, recommendations, and composable production APIs.
Create and refine images, video, and audio with scene, frame, timeline, layer, and export controls in one production workspace.
Open StudioTurn a prompt or ordered image list into a complete video project, then reroll individual scenes or continue editing in Studio.
Open VidGenieBrowse published generative videos in a searchable catalogue with related recommendations, topics, and responsive playback.
Browse GalleryTurn a topic into an interactive lesson with branching choices, alternate explanations, and a shareable learning path.
Open TMochiLearnRun or extend the generative video and media-automation platform in your own environment.
View repositoryAn A2A gateway for scoped Samsar accounts, credit purchase, text-to-video submission, status polling, and completed render retrieval.
View repositoryA Cloud Run-ready A2A gateway for agent registration, credit-backed access, authenticated video generation, and task polling.
View repositoryA TypeScript client library for integrating Samsar APIs into web applications, services, and automation workflows.
View repository04 / Approach
Each engagement covers the application, integration, data, and deployment work needed for a production release.
Define users, integrations, data, constraints, and a measurable first release.
Implement interfaces, services, agents, and pipelines in testable increments.
Containerize, configure, monitor, and iterate on the shipped system.
05 / Samsar journal
Updates on Samsar products, model integrations, generative workflows, search, recommendations, and deployment.
Samsar adds Alibaba Cloud model adapters for QWEN 3.7, WAN 2.7, and Happy-Horse 1.1, together with the Quinn A2A connector.
Read the updateDiscuss a software, agent, data, or deployment project.
roy@kashmerelabs.com