Résumé

Ketul Kishorbhai Chhaya

Germantown, MD (open to relocation)

AI-focused backend engineer. Architected BizChat, an LLM platform serving 220+ Baltimore entrepreneurs via the Mayor's Office, with routing across Bedrock/OpenAI/Anthropic and an open-source observability package.

Experience

Software Engineer

BizChat
May 2025 – Present
Baltimore, MD
  • Optimized Firestore read volume 108x (2,800 to 26 per query) by architecting an inverted index served through a Next.js API, with weekly incremental rebuilds on Cloud Run that skip unchanged records via content hash diffing.
  • Engineered zero-loss model switching for 220+ entrepreneurs by building a multi-provider LLM router across OpenAI, Claude, Mistral, and Llama with persistent session state in Firebase.
  • Reduced monthly LLM token spend 40% by replacing a custom Puppeteer scraper with the Tavily API, removing full page HTML from model inputs, traced via tokenwise-tracker, an open-source proxy I built.
  • Cut AI suggestion latency 72% (8s to 2.2s) by engineering a ProseMirror streaming pipeline in Next.js and TypeScript that renders NDJSON chunks as inline diff decorations with zero layout shift or document history pollution.
  • Quantified the user acceptance, revision, and retention patterns behind a CHI publication by piping 100K+ Firestore telemetry events into BigQuery and surfacing SQL results through an auto refreshing data connector.
  • Owned an end-to-end testing and release safety net across Chromium, Firefox, and WebKit, cutting QA triage time by 60% by automating a Playwright CI/CD pipeline with Slack alerts and hosted reports to catch pre-production regressions.

Software Engineer

Aavenir
Jan 2023 – Jun 2024
Ahmedabad, GJ
  • Overhauled contract migration from two weeks to five days across 8,000+ contracts by running AI metadata extraction in parallel with per document timeouts and fallback handling, built on Python ETL pipelines and Spring Boot REST services.
  • Developed a NestJS approval microservice that routes contracts through parallel review then sequential executive sign off by contract value and department, reducing legal approval cycle time 50%.
  • Prevented 6+ concurrent edit collisions on shared enterprise contracts by engineering real-time presence detection and distributed lock webhooks between Office.js Word add-ins and the core web app.
  • Automated new client instance provisioning with a Node.js toolkit that standardizes contract models, templates, fields, views, and business rules, saving 110+ engineering hours of manual setup across 4 deployments.
  • Decreased administrative setup overhead by 35% by developing an AngularJS administration console featuring drag-and-drop workflow builders, interactive audit logs, and form validation.

Full Stack Engineer Intern

ClosestCloset
Jun 2022 – Dec 2022
Remote (Chicago, IL)
  • Diagnosed a 5s MongoDB ranking bottleneck and cut query latency 84% to 800ms via automated score precomputation using Node.js based cron jobs and Redis caching.
  • Reduced customer support tickets by 40% by adding a buyer-seller chat portal using Node.js, Azure Web PubSub, and React with automated offline email fallbacks.
  • Unblocked the marketing operations team with a modular internal CMS in React/Redux featuring preview rendering, scheduling, and reusable UI components, removing 12 recurring developer support requests per week.

Projects

PayPipe|
FastAPIApache KafkaDockerPython
github.com/KetulChhaya/PayPipe

Engineered a fault-tolerant, event driven payment pipeline achieving exactly once processing via Redis SETNX idempotency keys, with exponential backoff retries and a dead-letter queue isolating failures.

Repo Graph|
TypeScriptts-morphSQLite
github.com/KetulChhaya/repo-graph

Built a TypeScript MCP server that parses a codebase into an import dependency graph, highlighting circular imports and tracing the blast radius of a file change across 1,448 files in 1.1s.

CiteSight – GEO Agent|
FastAPILangGraphTemporalpgvectorDocker
github.com/KetulChhaya/GEO-Agent

Orchestrated a LangGraph and Temporal pipeline that embeds a company site into pgvector, probes Claude, GPT, and Gemini with buyer intent questions, and scores brand visibility by measuring engine answers against the site's own content.

Scheduler Reflow|
TypeScriptLuxonVitest
github.com/KetulChhaya/scheduler-reflow

Implemented a TypeScript based scheduling engine that reflows orders around delays and maintenance windows via Kahn's topological sort and greedy shift-aware placement.

Technical Skills

Languages
PythonTypeScriptJavaScriptJavaSQLHTMLCSS
Frontend
ReactNext.jsAngularTailwindCSSProseMirrorReduxZustandWebpackTurborepoWCAGFigma
Backend
Node.jsNestJSExpressFastAPISpring BootREST APIsGraphQLPostgreSQLMongoDBFirestoreBigQuerySQLiteRedisKafkaTemporalWebSocketsCachingElasticSearch
Cloud / DevOps
AWSGCPAzureServiceNowDockerKubernetesGitHub ActionsCI/CDDatadogLinuxJUnitJestVitestPlaywright
AI / LLM
AWS BedrockOpenAIAnthropicLangChainLangGraphRAGMCP

Education

M.S. in Computer Science

University of Maryland, Baltimore County
Aug 2024 – May 2026
Baltimore, MD
GPA: 4.0/4.0·ranked #1 in cohort

B.Tech. in Information & Communication Technology

Pandit Deendayal Energy University
Aug 2019 – Dec 2022
Gandhinagar, GJ
GPA: 9.65/10·Top 5 in cohort