Commerce & Care AI
++ Architected CASA, an LLM-powered commerce assistant spanning POS, mobile, checkout, and catalog — designed for tens of millions of merchants, with multi-agent QA for continuous coverage. +
+diff --git a/.nojekyll b/.nojekyll new file mode 100644 index 0000000..e69de29 diff --git a/README.md b/README.md index 3cf6d9d..90df23d 100644 --- a/README.md +++ b/README.md @@ -1 +1,19 @@ # nitishagar.github.io + +Personal landing page for [Nitish Agarwal](https://github.com/nitishagar). + +Highlights merged open-source contributions, GoDaddy engineering leadership, selected projects, and recent writing on [Medium](https://medium.com/@nitishagar). + +## Local preview + +Open `index.html` in a browser, or serve the folder: + +```bash +python3 -m http.server 8000 +``` + +Then visit `http://localhost:8000`. + +## Deploy + +GitHub Pages serves from the `master` branch root of this repository. diff --git a/index.html b/index.html new file mode 100644 index 0000000..0d25504 --- /dev/null +++ b/index.html @@ -0,0 +1,322 @@ + + +
+ + +Principal Engineer · Open Source · AI Systems
++ Shipping production AI systems and merged contributions across the stack — while leading engineering teams that scale. +
+ +Leadership
++ Principal Engineer focused on AI transformation for Care and Commerce — building systems merchants and guides rely on every day. +
+ ++ Architected CASA, an LLM-powered commerce assistant spanning POS, mobile, checkout, and catalog — designed for tens of millions of merchants, with multi-agent QA for continuous coverage. +
++ Grew GoDaddy’s India engineering organization from a small core to 120+ engineers — hiring, org design, and delivery cadence that keeps platform work shipping. +
++ Leads Care Engineering’s CRM and tooling transformation for guides and customers — agent architecture, MCP, evaluation, and the operational discipline that survives production. +
+Merged contributions
++ Recent merged pull requests into projects other people run in production — CI, observability, LLM gateways, Kubernetes, and IoT platforms. +
+ +UI GraphQL fixes, gzip API middleware, docs, and reliability patches across Mozilla’s CI-at-scale platform.
+ View contributions + + + + +Chart cross-filter correctness, table sorting, report CSV export, dashboard ownership, and MySQL time-grain fixes.
+ View contributions + + + + +Proxy spend-flush correctness, premium metadata gating, and session-aggregate cost/duration in request logs.
+ View contributions + + + + +Chunked inline review comments into bounded batches for Alibaba’s hybrid deterministic + LLM code-review action.
+ View pull request + + + + +Enabled CONFIG_TASK_IO_ACCOUNTING and related kernel options in the x86_64 ISO for better I/O observability.
+ View pull request + + + + +Shell completion, dinit init support, and warnings when client discovery is disabled by loopback-only listen.
+ View contributions + +Repositories
++ Open tools for agent readiness, verifiable AI decisions, MCP authoring, and observability — the same themes as the writing. +
+ +Lighthouse reimagined for the AI-first world — audit AI agent readiness the way you audit web performance.
+Tamper-evident receipts for every AI decision — Ed25519-signed, hash-chained, offline-verifiable.
+Scaffold, grade, test, and debug MCP servers from inside Cursor — because broken tools are usually broken descriptions.
+Reversible proprietary-term pseudonymization for LiteLLM — mask codenames outbound, restore them faithfully inbound.
+MCP server for querying observability backends — OpenTelemetry Collector, Grafana, Elastic Cloud, OpenSearch.
+Agents that rerun your science and show their receipts — reproducibility auditors for scientific Python projects.
+Medium · @nitishagar
++ Field notes for engineering leaders — what survives production over what demos well. +
+ +AMD just shipped a $5 million rack. Nvidia sells factories. The AI industry has re-entered its System/360 era.
+ + + +Traditional monitoring assumes deterministic code. Agents aren’t. A field guide to observing and budgeting LLM agents.
+ + + +Agent memory is 2026’s vector database: real pain, serious funding — and a markdown file already doing most of the job.
+ + + +How the Agent Name Service went back to the drawing board — and why its second act matters to your architecture.
+ + + +Scientific Python solved environment specification. Verification still runs on volunteers — three agent patterns for the rest.
+ + + +The silent-failure problem nobody owns — and the Cursor plugin built to stop guessing.
+ +