whoami open source recently shipped how I build experience stack activity
I contribute to the Python and AI-infrastructure tools I build on — real fixes and features, each shipped with tests, in projects with active maintainer review.
A new container module that runs Google's fake-gcs-server emulator in tests, so GCS code can be integration-tested without cloud credentials. Modeled on the repo's existing Datastore pattern, with docs and tests included.
from testcontainers.community.gcs import GcsContainer
with GcsContainer() as gcs:
client = gcs.get_client()
bucket = client.create_bucket("test-bucket")
bucket.blob("hello.txt").upload_from_string("Hello from the emulator!")Why it holds up: a downstream project (panodata/omniload) swapped it in as a drop-in replacement for its own hand-rolled container — all 3 of its GCS integration cases passed unchanged, at the same runtime.
| Project | What I shipped | Status |
|---|---|---|
| testcontainers-python | New GcsContainer module (fake-gcs-server) with docs and tests |
Maintainer-approved, awaiting CI/merge |
| azure-search-openai-demo | Fix: results without a reranker score were treated as 0 against the default 1.9 threshold, dropping every result when the semantic ranker was unavailable | All 21 CI checks green, awaiting review |
| llama-agents | Root-cause fix for PickleSerializer returning wrong values for JSON payloads, with regression tests |
Open, awaiting review |
| langgraph | Fix in progress: Functional API resume re-runs a task whose exception was caught instead of replaying its recorded outcome | Claimed, building |
| Date | Work |
|---|---|
2026-10-28 |
Claimed langgraph #9475 — checkpoint task-replay bug in the Functional API |
2026-10-27 |
Opened llama-agents PR #838 — PickleSerializer round-trip fix (+203/−4, regression tests) |
2026-10-26 |
Opened testcontainers-python PR #1145 — new GcsContainer module (+262/−2) |
2026-10-23 |
Opened azure-search-openai-demo PR #3250 — reranker-score filtering fix |
flowchart LR
BUG["Bug report"] --> REPRO["Reproduce locally"]
REPRO --> ROOT["Find the root cause"]
ROOT --> FIX["Fix + regression tests"]
FIX --> PR["Pull request"]
PR --> REV["Maintainer review"]
REV --> MERGE["Merged"]
| Principle | What it looks like in practice |
|---|---|
| Root cause over patches | In llama-agents, the serializer bug had narrower fixes that only tightened validation; mine corrects the decode ordering that caused it, so the whole class of bad round-trips is covered. |
| Reproduce first | Every fix starts from the reporter's repro, run locally — no repro, no patch. |
| Tests ship with the fix | Regression tests land in the same PR (#838, #1145), so the bug stays fixed. |
| Fit the codebase | New modules follow the repo's own patterns — GcsContainer mirrors testcontainers' Datastore module — and PRs include docs, changelog entries and CLA housekeeping. |
Software Development Engineer · Quadrant Technologies (client: Microsoft) · Jun 2025 – present · Redmond, WA
- Built RAG-based AI agents on Azure AI Foundry with LangGraph and MCP, retrieving from Microsoft Fabric and Azure AI Search — cut manual lookup time 30% for the team using them.
- Own a reusable Python/Airflow framework running 70+ production workflows on PostgreSQL, MongoDB and cloud storage — design docs, design reviews, on-call, incident response.
- Drove C#/.NET compatibility tooling that catches Windows runtime regressions across 23 device configurations — runtime issues down 27% (300 → 219/month), root-caused via Event Logs and Win32 API traces.
- Automated DEV/UAT/PROD environment sync and database refresh — drift defects down 40% per release; SonarQube/Bandit quality gates and Grafana alerts cut deployment errors 25%.
Co-founder · Doppular
- Co-founded Doppular, a social app that matches you with people around the world who look like you — alike in looks, unique in life.
- Own the multi-tier distributed backend (Node.js, TypeScript, MongoDB, Redis, BullMQ) sustaining 100K+ API requests/day, plus a Python FastAPI + Milvus vector-search microservice running sub-200ms, on AWS with Docker and Kubernetes.
Software Engineer Intern · Skipline · May 2024 – May 2025 · Remote
- Owned telemetry ingestion services processing 1M+ daily records; validation cut ingestion failures 30%, and date-partitioned PySpark/SQL pipelines cut processing time 20%.
Research Assistant · Clarkson University · Sep 2023 – May 2024
- Built PyTorch/scikit-learn pipelines for satellite and UAV multispectral imagery (up to 98.4% balanced accuracy); published in Machine Learning with Applications.
Education: M.S. in Data Science, Clarkson University (GPA 3.8) · B.Tech in Computer Science, JNTUK
| Layer | Tools |
|---|---|
| Languages | Python · Java · C# · C++ · TypeScript · SQL |
| Backend & Data | REST APIs · Microservices · Kafka · Redis · PostgreSQL · MongoDB · Apache Airflow · Node.js · FastAPI |
| Cloud & DevOps | AWS (EC2, S3, Lambda, CloudFront) · Azure DevOps · Docker · Kubernetes · Grafana · CI/CD · Git |
| AI & Agents | RAG · AI Agents · MCP · LangGraph · Azure AI Foundry · Azure AI Search · GenAI |
Let's build systems that ship. Open to software engineering roles — backend, distributed systems, AI infrastructure.
