Nishtha Arora

Nishtha Arora

Senior Software Engineer | Frisco, TX

Senior Software Engineer with 8+ years building scalable frontend platforms and, more recently, AI-native developer tooling and agentic systems. Experience spans architecting shared React design systems, building production RAG pipelines and tool-calling agents on AWS Bedrock, and stepping in as acting manager, taking on sprint planning, cross-functional ownership, and mentoring.

Frontend EngineeringAI & Agent DevelopmentTeam Leadership
8+
years experience
84
components migrated
3+
teams served
85%
critical-path test coverage
4x
fewer deploy failures

nishtha.arora26@gmail.com · linkedin.com/in/nishthaarora

Experience

  1. Appfire, Remote, TX

    Senior Software Engineer · February 2025 – Present
    Team Leadership
    • Stepped in as acting manager, owning sprint planning, backlog grooming, and retrospectives, driving team execution rhythm and surfacing technical risk early
    • Serve as the primary cross-functional point of contact with product and design, aligning feature scope and roadmap phases with engineering capacity
    • Coach junior engineers through structured growth paths, helping translate mentoring into measurable progress toward more senior scope
    AI & Agent Development
    • Built the full RAG pipeline for the team's Flow AI agent: LangChain document ingestion and chunking, Amazon Titan embeddings via Bedrock, ChromaDB vector store, and semantic retrieval; evaluated retrieval quality across chunk sizes, embedding strategies, and similarity thresholds before shipping
    • Implemented the agentic backend in FastAPI with two-round-trip Bedrock tool calling (RAG retrieval + ticket-creation tool) and a FastMCP stdio server exposing the agent as an MCP-compatible service for Claude Code and other clients
    • Built a PR review agent on the Claude SDK running as a daemon triggered by GitHub notifications, leaving inline, recommendation-only comments; supports Claude, Gemini, and Cursor as drop-in AI providers
    • Built an AI Adoption Dashboard tracking GitHub Copilot and Cursor usage org-wide, giving leadership visibility into AI tooling adoption and ROI
    Frontend Engineering
    • Set React/TypeScript architecture patterns and led migration of 84 components into a unified design system (Flow), documented in Storybook; mentored 3 engineers
    • Streamlined CI/CD pipelines and ArgoCD workflows, cutting deployment failures from weekly to roughly once a month
    • Applied AWS/S3 to cloud infrastructure and data ingestion workflows, improving data availability

  2. Pluralsight, Remote

    Senior Software Engineer
    January 2024 – February 2025
    Frontend Developer
    September 2020 – January 2024
    • Architected and owned the shared React component library used across 3+ teams
    • Rolled out Datadog RUM to establish real performance baselines across the platform, shifting the team from guesswork to data-driven decisions
    • Wrote and championed Cypress integration tests, growing critical-path coverage from near zero to roughly 85%
    • Led a customer-facing onboarding revamp attributed with a ~15% lift in product adoption; cleared a backlog of 100+ bugs ahead of SLA deadlines

  3. Match Group, Dallas, TX

    Frontend Developer · June 2017 – September 2020
    • Built the user messaging system end-to-end using GraphQL (schema design, real-time subscriptions, React UI layer) for a consumer product with millions of daily users
    • Advanced A/B testing and analytics instrumentation across high-traffic user journeys; contributed to the org-wide internal design system

AI & Agent Projects

Nexus HR Agent

An AI-powered internal operations assistant built with AWS Bedrock, RAG, and FastAPI. Ask questions about company documentation or create support tickets — the agent routes your request to the right skill automatically. Built to learn and mirror production RAG + agentic patterns, alongside studying a production Strands Agents + S3 Vectors system at work to understand every architectural decision by building a working equivalent.
  • RAG pipeline over fictional company documentation: LangChain chunking, Amazon Titan embeddings via Bedrock, ChromaDB as the vector store — full document ingestion and semantic retrieval
  • Agent with tool calling: two-round-trip Bedrock format (initial call → tool invocation → second call with results). Two tools: search_docs (RAG retrieval) and create_ticket (support ticket creation)
  • Intent-based skill routing (documentation vs. support), stateful multi-turn conversations with session persistence, and conversation summarization for long sessions
  • FastAPI backend; Next.js + React + TypeScript + Tailwind chat frontend
  • FastMCP stdio server — exposes the agent as an MCP-compatible tool for use with Claude Code or any MCP client
  • LLM-as-judge evaluation pipeline for measuring retrieval and response quality

PR Deep Review Bot

A multi-agent deep PR reviewer that catches cross-file breakage, security vulnerabilities, convention violations, test gaps, and dependency issues that single-pass reviewers miss. Built with claude-agent-sdk as a 3-phase pipeline: triage, parallel specialist analysis, and synthesis.
  • 3-phase pipeline: Triage (haiku, classifies risk and picks specialists) → Specialists (sonnet, run in parallel: cross-file breakage, security, architecture, test gaps, dependencies) → Synthesis (haiku, dedups and formats findings)
  • Confidence scoring: every finding gets a 0-100 score, with configurable per-category thresholds to filter noise
  • Dedup against existing PR comments using semantic matching, not just string comparison
  • Per-repo learnings stored in JSON — specialists receive past findings as context to avoid re-flagging known issues
  • GitHub Action triggers: on-demand via PR comment, or automatic on PR open/synchronize above a configurable changed-files threshold
  • External, versioned prompt templates — editable without code changes

PR Review Bot

A fully local AI PR review bot powered by Ollama — reviews pull requests from any GitHub repository without sending code to a cloud API. Built to explore what a private, self-hosted code review agent looks like end to end.
  • Runs entirely locally via Ollama (deepseek-coder / qwen2.5-coder) — no code leaves the machine
  • Priority scoring across severity tiers (critical/high/medium/low/style) with configurable minimum-severity filtering to reduce noise
  • Parallel file processing and incremental review support for faster turnaround on large PRs
  • Streaming responses from the model as review comments are generated
  • Simple CLI: point it at any owner/repo and PR number to review

Skills

Leadership

MentoringSprint Planning & GroomingCross-Functional CommunicationGrowth CoachingArchitecture StandardsAgile/SDLC

AI & Automation

AWS Bedrock (Claude Sonnet 4.6, Titan)RAG PipelinesAgentic Tool CallingMCP / FastMCPChromaDBLangChainClaude SDKPrompt EngineeringLLM Evaluation

Frontend

ReactTypeScriptJavaScriptNext.jsReduxGraphQLHTML5CSS3

Testing & Observability

CypressPlaywrightReact Testing LibraryJestDatadog RUMStorybook

Platform

Node.jsPythonDjangoFastAPIPostgreSQLAWS (S3, Lambda, Bedrock)DockerArgoCDGitHub ActionsSQL

Education

  • Coding Bootcamp, Computer Programming, The University of Texas at Austin (2016–2017) — HTML, CSS, JavaScript, Node.js, Express, React, Angular, MongoDB
  • MBA, Finance & Marketing, Institute of Marketing and Management, India (2008–2010)
  • BBA, IP University, India (2005–2008)