Nikhil Yarram
Software Engineer
Backend engineer at Chewy building event-driven order systems on AWS. On the side: AI evaluation tooling and full-stack products in Python and TypeScript.
About
I'm a backend engineer at Chewy, working on the order platform: event-driven systems on AWS, payment failure recovery, and split-fulfillment logic.
Outside of work I build things I'd want to use. AstroDigest is a full-stack product that turns astronomy news into a weekly digest, and Litmus is an open-source harness that catches quality regressions in LLM pipelines before they ship.
I studied at the University of Washington and the University of Wisconsin–Madison. I'm based in Minneapolis and relocating to the Bay Area.
Experience
Chewy
Apr 2025 – PresentSoftware Engineer
Minneapolis, MN
- ▸Architected partitioned event hydration, migrating 50 production subscriptions to V2 schemas without coordinated topic cutovers
- ▸Built async degraded-payment recovery (SQS/SNS) with idempotent reconciliation, preserving order intake during payment outages
- ▸Implemented allocation-aware split-fulfillment logic, piloted across 180 production orders spanning inventory, tax, and fulfillment
- ▸Extended delivery instructions across APIs, persistence, and events for an initiative projected to save $3.6M–$7.7M annually
- ▸Refactored shipment notification routing to a unified handler, retiring legacy code and removing 994 lines of dead logic
Indigo Systems & Technology Consulting Inc
Sept 2023 – Apr 2025Software Engineer
Chicago, IL
- ▸Architected a Kafka-integrated log analysis backend with Java Spring Boot & OpenSearch, processing 100K+ daily log entries and improving log query efficiency
- ▸Developed a React dashboard on a Node.js server, enabling real-time authentication tracking and log visualization, reducing debugging time by 50%
- ▸Engineered custom Java authentication nodes for Forgerock AM, facilitating risk-based authentication and session state management through shared-state logic
- ▸Designed a high-throughput, low-latency CRUD API on Docker, supporting 1M+ monthly identity operations across multi-team microservices
- ▸Automated ForgeRock AM deployments via Kubernetes and Helm, streamlining GCP-hosted upgrades, rollbacks, and multi-environment deployments
- ▸Developing a Python-based fraud detection system, integrating IP analysis, login patterns, and anomaly detection, reducing false-positive security flags by 10%
Oshkosh Corporation
Jun 2022 – May 2023Co-op Software Engineer
Oshkosh, WI
- ▸Crafted a fleet management platform in Python, enabling real-time vehicle telemetry and predictive maintenance to reduce fleet costs by 25%
- ▸Implemented AWS S3 & EC2 for scalable data storage, lowering infrastructure costs by 15% while enhancing data accessibility
- ▸Optimized PostgreSQL indexing strategies, improving query performance and real-time analytics by 20%
- ▸Built RESTful APIs to integrate vehicle logs, streamlining fleet data processing and real-time monitoring
University of Wisconsin, Department of IT (DoIT)
Oct 2021 – May 2022Identity and Access Management (IAM) Intern
Madison, WI
- ▸Enhanced MFA reliability by implementing secure access solutions in C++, reducing unauthorized access incidents by 15%
- ▸Automated identity data management tasks using SQL and Azure, improving data accuracy and efficiency by 20%
- ▸Increased system reliability by 30% through MSTest-driven security API validation
- ▸Contributed to Agile development in a Scrum environment, participating in biweekly sprints, sprint planning, and daily standups
Projects
A weekly AI-curated digest of astronomy news and research, delivered to web and mobile. Content from NASA, ESO, ALMA and arXiv is ingested daily, scored and summarized by an LLM, then assembled into one digest.
- ▸Designed a monorepo with web, mobile and API apps plus five packages for ingestion, scoring, summarization, editorial review and delivery
- ▸Split the pipeline across Cloudflare Workers (daily ingestion cron) and a Dockerized BullMQ worker fleet with self-hosted Redis
- ▸Added an editorial worker that flags low-quality and refusal summaries before they reach a digest
- ▸Deploys to a VPS on every merge to main through GitHub Actions; web ships on Vercel
An open-source evaluation harness for LLM pipelines and agents. It answers "did my last change make the system better or worse?" and fails CI when quality regresses.
- ▸Runs versioned golden datasets through deterministic scorers and LLM judges, then diffs each run against a committed baseline
- ▸Calibrates judges against human labels before trusting them, and gates PRs on a score threshold
- ▸Traces every score back to the model call that produced it with OpenTelemetry GenAI conventions
- ▸Runs on free tiers or local models; 41 tests need no API keys
A sample-tracking and inventory backend for a regulated laboratory, built on the idea that invariants belong in the database, not the application.
- ▸Enforces sample lifecycles, non-negative stock and storage occupancy with constraints and triggers, so no writer can violate them
- ▸Writes a tamper-evident, append-only audit trail with database triggers and a hash chain, covering shell sessions and DBAs as well as the API
- ▸Implements FEFO inventory allocation and reservations; 64 tests run against real PostgreSQL
Skills
Languages
Backend & Data
Cloud & Infrastructure
Frontend & ML
Movies I Love
Passionate movie buff! Always looking for the next one to rave about. The medium that transcends cultural barriers!
Get In Touch
I'm always open to discussing new opportunities, interesting projects, or just a good conversation about software. Drop me a line.
Based in Minneapolis, MN · Relocating to the Bay Area