Skip to content
View AbdullahRasheed45's full-sized avatar

Block or report AbdullahRasheed45

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.

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
AbdullahRasheed45/README.md

Muhammad Abdullah Rasheed

ML Research Engineer · scientific computing, neural operators, and applied machine learning

LinkedIn Email Website


About

ML Research Engineer at Aeris UK, based in London. MSc Data Science (Distinction). I work at the intersection of numerical methods and machine learning — reinforcement learning and simulation software, PDE solvers and neural-operator surrogates, distributed training pipelines, and LLM-driven agents. I care about code that is correct, tested, and honest about its trade-offs: most of my recent projects ship with a validation suite, CI, and benchmarks rather than just a notebook.

  • Currently building reinforcement-learning and simulation software at Aeris UK.
  • Research interest in operator learning for PDEs and high-performance numerical computing.
  • Comfortable across the stack: C++/OpenMP kernels, PyTorch (DDP/FSDP), and full-stack deployment.
  • 15+ ML models deployed across research and production settings.
  • Reach me at abdullahrasheed45@gmail.com or on LinkedIn.

Open source contributions

Repo Merged pull request(s)
huggingface/transformers (163k⭐) #47663 fix: correct text input validation logic in 8 multimodal processors (and → or)

Also have 19 pull requests currently under review across 13 other repositories:

Open pull requests
Repo Pull request(s)
stanfordnlp/dspy (37k⭐) #10112 fix: deduplicate signatures by content in get_dspy_source_code (fixes silent drop of dynamic signatures) · #10085 Roll nested track_usage() scopes up into the enclosing tracker
Lightning-AI/pytorch-lightning (31k⭐) #21866 Bump rich upper bound to <15.1.0 to support rich 15.0
pyg-team/pytorch_geometric (24k⭐) #10762 validate(): raise on numpy arrays stored as tensor attributes · #10756 Verify TLS certificates when downloading datasets
huggingface/datasets (22k⭐) #8368 Reject duplicate column names in Dataset.select_columns
huggingface/peft (21k⭐) #3477 Remove PEFT forward hooks when the wrapped block raises
onnx/onnx (21k⭐) #8215 Fix node test directory count in generate-data
EleutherAI/lm-evaluation-harness (13k⭐) #3971 fix: prevent ValueError when batch_size="auto" is passed to neuronx model · #3970 fix: prevent ValueError when batch_size="auto:N" is passed to API models · #3951 Include generation_kwargs in the request cache key
facebookresearch/hydra (11k⭐) #3333 fix(optuna-sweeper): pass values=None when marking trial as FAIL, add NaN check
huggingface/accelerate (9.8k⭐) #4138 Reject sharded-checkpoint shard paths that escape the checkpoint folder
bentoml/BentoML (8.7k⭐) #5676 Keep every URL value of a multipart list field
cvxpy/cvxpy (6.3k⭐) #3470 Stop internal parameter checks warning about sparse .value reads
huggingface/lighteval (2.5k⭐) #1314 Fix always-true conditional that made DEFAULT_FORMAT unreachable · #1313 Fix litellm judge sending max_tokens as a 1-tuple
Lightning-AI/torchmetrics (2.5k⭐) #3441 fix(validation): correct argument validation in auroc, logauc, and pesq metrics · #3440 fix(validation): correct 'and' to 'or' in argument validation across multiple metrics

Featured work

Project What it is Stack
fno-darcy-flow Fourier Neural Operator surrogate for 2D Darcy flow, with a from-scratch finite-volume data generator and 1-vs-N-GPU scaling benchmarks. Trains identically under DDP and FSDP/ZeRO-3. PyTorch, DDP/FSDP, NumPy
darcy-cpp Matrix-free multigrid finite-volume solver for 2D Darcy flow. Grid-independent CG iteration count, OpenMP-parallel, with zero-copy Python bindings — up to 16× faster than a sparse direct solve. C++17, OpenMP, pybind11
Adaptive-Micro-Gesture-Recognition Master's thesis: a 3D CNN that recognises subtle hand movements as computer commands to improve accessibility for motor-impaired users (94.6% accuracy). TensorFlow, OpenCV, Flask
Grammify A T5 Seq2Seq grammar-correction app with real-time error analysis, deployed on Hugging Face Spaces. Transformers, FastAPI, Streamlit
crypto_trading_ai_agents Agentic system that forecasts price movements and produces buy/sell signals. Python, LLM agents

Tech stack

Languages · Python, C++, SQL, JavaScript/TypeScript, R

ML / scientific · PyTorch TensorFlow NumPy SciPy scikit-learn LangChain

Systems / tooling · C++ CMake Docker GitHub Actions PostgreSQL AWS

Web · React Next.js Node.js

Experience & education

  • ML Research Engineer — Aeris UK (current) · reinforcement learning and simulation software
  • Data Analyst — Digital Pulse 360
  • Full-Stack Developer — ATnR, Karachi
  • MSc Data Science — Anglia Ruskin University, Cambridge (Distinction)
  • BSc Economics — Institute of Business Administration (IBA), Karachi (Dean's Honour List)

Certifications · Google Advanced Data Analytics · Google AI Essentials · Google Data Analytics Professional · Tableau Specialization (UC Davis)

Pinned Loading

  1. Adaptive-Micro-Gesture-Recognition Adaptive-Micro-Gesture-Recognition Public

    Python

  2. ai-agent-github-mcp-agent ai-agent-github-mcp-agent Public

    Python

  3. ai-projects-deep-researcher-agent ai-projects-deep-researcher-agent Public

    Multi-stage deep researcher agent using Agno and ScrapeGraph with Streamlit UI.

    Python 1

  4. GENERATIVE_AI_MEDICAL_ASSISTANT GENERATIVE_AI_MEDICAL_ASSISTANT Public

    Jupyter Notebook

  5. Movie_sentiment_analysis_MLOPS Movie_sentiment_analysis_MLOPS Public

    This is the machine learning program that follows a complete mlops workflow and frames.

    Python