DevOps · Developer Experience · AI Memory Systems
Ayushman Bhattacharya
DevOps, Platform & Developer Experience
I build the CI/CD, release, observability, contributor automation, and developer tooling that help open-source and AI systems remain understandable and operable.

One practice, three connected lanes
The work is organized by what it proves.
My primary path is DevOps and platform engineering through developer experience: CI/CD, release systems, observability, cloud deployment, contributor automation, and safe self-service. My AI infrastructure specialization is memory management, including caching, context assembly, retrieval, invalidation, privacy, and recomputation. Developer Relations is how I document and teach work after it exists.
DevOps, platform and Developer Experience
CI/CD, release engineering, service health, contributor automation, access boundaries, and the systems that let other developers ship and operate software safely.
Explore this lane 02 / Technical minorMLOps memory and inference systems
Caching, context assembly, retrieval, invalidation, privacy, recomputation, and observability for open AI services—grounded in reproducible research and running systems.
Explore this lane 03 / Public practiceOpen-source DevRel and maintainership
Contributor paths, documentation, technical talks, and community programmes derived from implementation rather than separated from it.
Explore this laneSelected technical work
Systems with a public record.
pollinations
An open-source generative AI platform where I maintain contributor rewards, triage automation, application workflows, model curation, and developer tooling.
Open repositoryCrumb
A self-hosted Rust terminal agent with explicit permissions, isolated tools, supervised MCP execution, signed skill distribution, rollback, and revocation. Debian packaging is in progress on Salsa (ITP #1151280); its execution-receipt workflow remains under validation.
Open repositoryOreoLook
An OpenAI-compatible web-search and research system with Redis-backed conversational memory, semantic caching, retrieval, MCP tools, streaming responses, citations, and reproducible evaluation artifacts.
Open repositoryElixpo Packages
The source, release, signed APT distribution, and public catalogue for independently versioned Elixpo Linux applications, with candidate deployments and rollback documentation. Selected for the XERJ Corpus Hub (in licence review).
Open repositoryElixpoo
Elixpo's organization-wide GitHub operations agent for deterministic triage, Project V2 automation, permission-checked metadata actions, comment-only reviews, and merge records.
Open repositoryElixpo Status
A Cloudflare-native status surface that derives service health from observed traffic, distinguishes outages from unavailable monitoring, and publishes operational history across Elixpo.
Open repositoryQFabric
A research runtime for measured recovery across Linux and RTOS execution domains, with timing contracts, safe switching, replayable decisions, probation, and rollback.
Open repositoryLixBlogs
An operated, community-developed publishing platform with real-time collaborative editing, multi-tenant publications, Markdown, SEO, and API and CLI automation on Cloudflare.
Open repositoryLixSketch
An open SVG whiteboard engine and React canvas with a hand-drawn visual system for sketches, diagrams, documents, and presentations.
Open repositoryElixpo Accounts
Shared OAuth 2.0 and OpenID Connect identity infrastructure for Elixpo services, with SSO, passkeys, webhooks, and an edge-safe TypeScript SDK.
Open repositoryEvidence, not a claim wall
Every role should connect to work someone can inspect.
Projects link to source. Research links to its record. Talks follow implemented lessons. Experience is described through responsibility rather than title alone.
How the work developedResearch and public explanation
Build first. Then publish the lesson.
QFabric: Guarded Cross-Domain Task Recovery and Its Limits on a Linux-RTOS Board
arXiv preprint · arXiv:2611.01598
Read the record Research / 2026RV565: Cross-Layer Co-Design for Frame-Independent Video Playback on MicroPython-Class ESP32-S3 Systems
arXiv preprint · arXiv:2610.35935
Read the record Talk / 24 October 2026Why Hacktoberfest Stopped Counting Pull Requests in the Age of AI Agents: A Maintainer's View
React Kolkata Meetup: Oct '26 Edition
View the session Talk / 23rd August 2026The Client Boundary - What 'use client' Really Does in Modern React.
Techno India University
View the sessionWork together