Skip to content
View gratycodes's full-sized avatar

Block or report gratycodes

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
gratycodes/README.md

Gratus Antin

Senior Software Engineer at PayPal, based in Chennai, India, with about 11 years of experience. I build agentic AI systems and LLM tooling, Java/Spring Boot and Kafka backends, payments platforms, and Next.js frontends. Before PayPal I was a Principal Software Engineer at Oracle, and earlier worked at Citi, DTCC, Cognizant and Hexaware. The projects below are personal projects.

Stack: Java, Spring Boot, Kafka, Python, Go, TypeScript, Node.js, React, Next.js, LLM and agent tooling.

Projects

Agents and AI observability

  • dispute-copilot: a chargeback triage system on synthetic data, where an agent calls read-only MCP tools to recommend a decision with cited evidence and a reviewer approves or overrides it in a React console, backed by a Spring Boot service with Postgres and Kafka.
  • agent-runtime-kit: a minimal Python agent loop with a tool registry, optional OpenTelemetry GenAI tracing, and record/replay for regression tests.
  • otel-genai-cost: an OpenTelemetry Collector processor that prices GenAI spans in USD, records spend by model and custom dimensions, and alerts on daily budgets.
  • genai-trace-diff: a browser-only React app that turns GenAI traces into a chat timeline and diffs two runs (live demo).
  • kafka-ops-mcp: a Java/Spring Boot MCP server for safe Kafka and DLQ inspection, with dry-run-gated replay.

Frontend

  • chat-fault-lab: a Next.js app that injects stream failures (stalls, dropped connections, 429s, broken frames, split UTF-8, streamed tool calls) into an OpenAI-compatible chat and shows an accessible UI handling each one, with Playwright and axe tests (live demo).

Backend and payments

Platform and DevOps

  • flipfactor: re-runs a flaky test while varying one factor at a time (order, timezone, locale, seed, clock, CPU, parallelism) and reports which one flips the result, for pytest, Jest and go test, with a GitHub Action.
  • kube-rightsize-diff: a Go PR check that compares Kubernetes CPU and memory requests with Prometheus usage and comments with cost impact.
  • schema-evolution-gate: a Go CLI and GitHub Action that fails PRs breaking Kafka event schemas (Avro, JSON Schema) using Schema Registry compatibility rules and sample replay, no registry needed.
  • tfplan-triage: a Go CLI and GitHub Action that reviews Terraform/OpenTofu plans in one PR comment, ranking destroyed data stores, IAM widening, new internet exposure and drift across AWS, GCP and Azure, with label-based approvals.

ML, evals and data

  • fraud-feature-pipeline: a fraud-scoring pipeline on synthetic card data with shared streaming and offline features (Kafka, Redis, DuckDB), a LightGBM model with leakage-safe training, a FastAPI scoring API, and train/serve skew and drift checks.
  • evaldelta: Python stats for LLM evals (bootstrap CIs, McNemar, power, judge agreement) and a GitHub Action that posts the result on a PR.
  • rag-evaluation-lab: a small Python harness that scores RAG output for recall@k, MRR, citation coverage and lexical groundedness.

LinkedIn

Popular repositories Loading

  1. payment-ledger-platform payment-ledger-platform Public

    Spring Boot double-entry ledger API with idempotent posting and per-currency balance checks.

    Java

  2. rag-evaluation-lab rag-evaluation-lab Public

    Dependency-free Python harness that scores RAG output: recall@k, MRR, citation coverage and groundedness.

    Python

  3. agent-runtime-kit agent-runtime-kit Public

    Minimal Python agent loop with a tool registry, pluggable LLM, transcript memory and golden evals.

    Python

  4. gratycodes gratycodes Public

    Profile README

  5. otel-genai-cost otel-genai-cost Public

    OpenTelemetry Collector processor that prices GenAI spans in USD from token usage, with cost metrics by model/feature/tenant and daily budget alerts

    Go

  6. genai-trace-diff genai-trace-diff Public

    Read and diff OpenTelemetry GenAI agent traces in the browser: chat-style timeline, tokens, cost and latency per step, side-by-side run diff. No backend.

    TypeScript