Harsh Sharma

Building & learning

Tech

I build distributed backend platforms where a small visible action depends on a long chain underneath it. My work has crossed APIs, queues, enrichment and aggregation workers, template systems, databases, autoscaling, observability, and multi-environment delivery.

I work mainly across TypeScript, Node.js, and Java, with React and GraphQL experience that lets me follow the product path end to end. I also build AI-native workflows for the engineering lifecycle, from understanding a ticket and repository through planning, implementation, review, merge-request creation, and testing.

Experience

Platform work at AlphaSense

From product-facing React and GraphQL work to distributed services, production rollouts, and platform ownership.

  1. 2024 — now

    India

    Software Engineer 2

    AlphaSense

    I build distributed, event-driven platform services for alerting and multi-channel notification delivery.

    In H1 2026, I took end-to-end architectural ownership of a major alert-delivery migration. It reduced measured end-to-end p95 latency from 3.04 seconds to 1.41 seconds, while per-pod efficiency on the shared delivery path reached 5–6× its previous level.

    The backend work spans Java, TypeScript, Node.js, message queues, Kubernetes, Helm, ArgoCD, and KEDA. I also plan staged rollouts and rollback paths, strengthen release validation, and investigate production behavior across application and infrastructure layers.

    I built React micro-app experiences and GraphQL-backed product workflows. That full-stack foundation now helps me reason about a feature from user interaction to backend delivery and production behavior.

AI-native engineering

AI across the full development loop

I use AI across the lifecycle of engineering work and keep human review, testing, and verification visible.

I build reusable skills for recurring engineering problems. One example is an email-compatible HTML workflow that I applied to production template work and refined from the rendering failures I encountered.

I also add repository context through AGENTS.md, CLAUDE.md, architecture notes, domain concepts, and testing guidance so coding agents can navigate unfamiliar services more safely. A focused migration skill grew into an extensible plugin and then into dynamic workflows inside my personal engineering harness.

The working loop

  1. 01Ticket and repository context
  2. 02Analysis and implementation plan
  3. 03Code changes and review
  4. 04Merge request and testing
  5. 05Human verification and delivery

Currently building

Personal engineering harness

I’m turning the AI workflows I use for repository context, planning, implementation, review, merge-request creation, and testing into a personal engineering harness. The public version will keep human review and verification visible.

In progress

Writing

Engineering notes

What I understand better after trying, debugging, and revisiting.

I’m preparing notes from production lessons. They will appear here after the examples and claims can stand on their own.

Inspirations

The references behind the work

Technical books, essays, talks, and conversations live in the Library.

Open Library ↗