Software Engineer | Backend & AI | Open-Source Contributor

I'm a Data Science graduate interested in backend engineering, AI/ML, and building reliable software. I enjoy solving problems, improving existing codebases, writing tests, and contributing to open-source projects.

• Languages: Python, C++, Java, JavaScript, TypeScript, PHP, Rust
• Backend & APIs: FastAPI, REST APIs, microservices, gRPC
• Databases: PostgreSQL, SQL, DuckDB, Firebase
• AI/ML: Machine Learning, RAG, LLM applications, embeddings, AI agents
• Frontend: React, TypeScript, Vue
• DevOps & Systems: Docker, Linux, Git, GitHub Actions, CI/CD
• Testing & Code Quality: pytest, GoogleTest, regression testing, static analysis
• Protocols & Integration: MCP, A2A, WebAssembly (WASM)

• 🔵 Contributing to Canonical and Ubuntu projects
• 🔵 Working on Docling and document-processing pipelines
• 🔵 Contributing to projects across IBM, Google, and other open-source communities
• 🔵 Working with SoftTrack
• 🤖 Building Notescape, an AI-powered learning workspace
• 🐧 Exploring Linux, containers, distributed systems, and systems programming

_{46 merged and 49 open pull requests across 18 upstream projects I don't own. Each count links to the full list on GitHub.}

Project	Stars	Merged	Open	Latest PRs
docling docling-project · Document parsing for gen AI		8	4	• docs: document installing docling-slim without RapidOCR • fix(jats): convert the content of sub-articles • fix(md): keep a heading or list item without text from taking a later b… and 5 more
guava google · Google core libraries for Java		2	2	• Treat leading and trailing $ in a class name as part of a top-level c… via Copybara • Document that SimpleStatsCounter.snapshot() may be inconsistent with… via Copybara
styleguide google · Google style guides		8	5	• js: add an example for annotating empty collections • js: add a template literal example for long strings • js: add an example for the destructuring tip and 5 more
OpenMetadata open-metadata · Metadata platform for data and AI		2	3	• Fixes 35685: Build the profiler ORM for tables with nested array columns • Fixes 34856: Parse run_pipeline trigger response as PipelineServiceClie…
multipass canonical · Ubuntu VMs on demand		1	4	• fix(hcs): enable Hyper-V time synchronization
lxd canonical · System container and VM manager		13	3	• snap: Fix parsing of MicroCeph/MicroOVN units for shutdown ordering • Network: Report clear error for stale mutual peering • Resources: Report ethtool link modes above 100G and 10 more
foundations cloudflare · Rust service foundations library		2	1	• fix(telemetry): accept memory profiler settings on all platforms • feat(telemetry): expose a check for collisions with built-in server rou…
docling-pipelines docling-project · Scalable ingestion pipelines		9	2	• fix(extract): fail Docling Serve v2 extraction without primary content • feat(storage): share one PostgreSQL engine per connection configuration • feat(ingest): support file events in the OneDrive ingest provider and 6 more
ubuntu-lint ubuntu · Ubuntu packaging linter		1	—	• dput: support source format 1.0 packages
react react · Library for web and native UIs		—	2	• [DevTools] Stop polling the backend manager from the content script open • [DevTools] Release host instances deleted from a still-mounted component open
grpc grpc · RPC framework		—	4	• [Bazel][Python] Fail early when a .proto path contains '-' open • [Python] Implement the List RPC in the health checking servicers open • [Python Otel] Do not hold the census buffer mutex while exporting data open and 1 more
googletest google · C++ testing framework		—	1	• Fix Visual Studio header files for gtest open
qwen-code QwenLM · AI coding agent for the terminal		—	3	• fix(core): fall back on a 429 that carries an unrecognized provider code open • fix(cli): accept a server name in qwen mcp reconnect open • chore(deps): update vulnerable dependencies open
mcp-context-forge IBM · MCP gateway and registry		—	3	• feat(catalog): add Parlor.sh MCP server open • fix(a2a): normalize file parts between protocol versions open • fix(a2a): use hyphenated v0.3 task states open
cloud-init canonical · Cloud instance initialization		—	2	• fix(netplan): do not enable DHCPv6 for ipv6_slaac subnets open • feat(nocloud): support platform variables in seed URLs open
snapcraft canonical · Snap package build tool		—	1	• feat: add snapd-info project key open
domjudge DOMjudge · Programming contest jury system		—	3	• Don't let the browser restore jury toggle state on reload open • Fix double counting of open-to-all contests on jury teams page open • fix(resolver): log delete event when disabling team open
docling-core docling-project · Docling document model and serializers		—	6	• fix(file): let a proxy resolve the hostnames of proxied URL sources whe… open • fix(doc): drop references to cascade-deleted items and comments in dele… open • fix(html): nest a sub-list parented by the list group inside a list item open and 3 more

Rust-based HTML → Markdown converter with WebAssembly bindings. My work was reviewed by the project maintainer and code owner, then landed through follow-up PRs with my commits and authorship preserved.

• #720 — Preserve block separation in table cells · fixes #692
  Block-level elements inside table cells no longer join the surrounding text. Added <center>, <dialog>, <hgroup> and <search> as block-level elements, with regression tests for the Tier 1 and Tier 2 converters and different table configurations. Integrated through #733.
• #721 — Preserve WASM visitor handle ownership · fixes #517
  Assigning a WasmVisitorHandle no longer consumes the JavaScript handle, so it can be reused. Updated the Alef-generated bindings and added regression tests for visitor reuse and clearVisitor(). Split into #726 (Alef upgrade and regenerated bindings) and #727 (tests and WASM coverage).
• #652 — Preserve blockquotes in task items
  Initial fix and regression tests for a blockquote at the start of a task item being rendered on the checkbox line. The underlying bug was fixed on the integration branch by another PR; the maintainer kept my tests with me as commit author, and they reached main through #581.
• #701 / #703 — Keep a <br> on its own line as a hard break
  Original fix for a <br> on its own source line becoming a paragraph break, plus regression tests for block elements, comments, whitespace, inline wrappers and newline styles. Carried forward into #717, where the maintainer extended it to the fast converter and more edge cases; merged with my authorship preserved.

Technologies: Rust · WebAssembly · HTML/Markdown conversion · Automated testing

• Languages: Python, Rust, C++, JavaScript, TypeScript, PHP
• Backend & Databases: FastAPI, PostgreSQL, SQL, DuckDB
• Testing & Quality: pytest, GoogleTest, Ruff, regression testing, coverage enforcement
• AI & Document Processing: Docling, entity extraction, document parsing, RAG
• Protocols & Integration: gRPC, AsyncIO, MCP, A2A, WebAssembly
• DevOps & Systems: Docker, Linux, Git, GitHub Actions, Multipass, Hyper-V, cloud-init, Snapcraft
• Performance & Security: Lazy imports, logging optimization, dependency updates, large-graph performance

Notescape — AI-Powered Learning Workspace

An AI learning platform designed to help learners understand study material, identify weak topics, and plan revision.

• 📚 Document ingestion with OCR and support for multiple file formats
• 🧠 Retrieval-Augmented Generation (RAG), embeddings, and LLM-powered assistance
• 📝 Flashcards, quizzes and spaced repetition
• ⚙️ Stack: React, TypeScript, FastAPI, PostgreSQL, pgvector, Redis, Docker

• Competitive programming and algorithmic problem solving
• Codeforces and LeetCode practice
• Experience as a teaching assistant and competitive-programming mentor

• 100+ contributions in the last year (previously tracked; see GitHub for the current count)
• Open-source contributions to Canonical, Docling, Google, Cloudflare, OpenMetadata and more
• Contributions spanning backend development, testing, performance, AI, and systems engineering

• LinkedIn: Lubaba Khalid

Always learning, building, and contributing to open source. 🚀