Skip to content
View mohsinm-dev's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report mohsinm-dev

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 supported. This note will be visible to only you.
Report abuse

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

Report abuse
mohsinm-dev/README.md

Mohsin Mehmood

Machine Learning Engineer · LLM Systems & Agent Architectures · Open-Source Contributor

Forward Deployed ML Engineer at Kodamai (Glasgow). I design and build production core ML and AI systems: multi-agent pipelines, document extraction infrastructure, and LLM-powered workflow automation. Most of my work sits at the boundary between production-grade ML and the messy reality of deploying it in regulated, high-stakes environments.

I carefully consider system architecture before writing code, including transaction boundaries, failure modes, inference latency budgets, and how models degrade in production. I have shipped AI systems in healthcare (HIPAA/PDPA), enterprise supply chain, and early-stage startups, working directly with non-technical domain experts to translate complex processes into automated pipelines.


Open-Source Contributions

Project Contribution
google/flax #5 contributor (last 12 months, official leaderboard · 19 commits). Merged PRs in Flax NNX core: fixed nnx.tabulate crash with empty/None values (#4891); fixed variable hook display bugs in nnx.tabulate (#5008). Both shipped in v0.12.1.
python/cpython 6 merged PRs: fixed TextIOWrapper.tell() assertion failure with standalone carriage return (backported to 3.13 + 3.14 · #141331); fixed re.Match.group() doc claiming [1..99] range limit (#144696); documented asyncio Task cancellation propagation (#141249); fixed inaccurate object comparison docs (#141221).
pytorch/pytorch Approved PR in PyTorch core (#167209).
uber/causalml Fixed bug where estimation_sample_size was not propagated from UpliftRandomForestClassifier to individual trees (#850). Silently broken behaviour in the core uplift Cython layer.

Technical Depth

core_ml = {
    "frameworks":     ["PyTorch", "JAX/Flax", "TensorFlow"],
    "training":       ["LoRA", "QLoRA", "RLHF", "DPO", "mixed precision", "distributed training"],
    "inference":      ["quantization (GPTQ/AWQ/bitsandbytes)", "vLLM", "GGUF/MLX", "TensorRT"],
    "evaluation":     ["RAGAS", "ROUGE", "BERTScore", "bootstrapped CI", "MLflow", "W&B"],
    "architectures":  ["Transformers", "CNNs", "multi-agent systems", "RAG pipelines"],
}

systems = {
    "design":         ["modular monolith", "worker architecture", "state machines", "transactional outbox"],
    "infra":          ["Docker", "Kubernetes", "FastAPI", "Redis", "RabbitMQ", "Celery"],
    "cloud":          ["GCP Vertex AI", "AWS SageMaker / EC2 / Lambda", "Azure ML"],
    "observability":  ["structured logging", "latency tracing", "drift monitoring", "PHI/PII redaction"],
    "compliance":     ["HIPAA", "Singapore PDPA"],
    "languages":      ["Python", "C++", "Bash"],
}

Contact

Pinned Loading

  1. flax flax Public

    Forked from google/flax

    Flax is a neural network library for JAX that is designed for flexibility.

    Jupyter Notebook

  2. augur augur Public

    Forked from chaoss/augur

    Python library and web service for Open Source Software Health and Sustainability metrics & data collection. You can find our documentation and new contributor information easily here: https://oss-…

    Python

  3. openbiollm-inference openbiollm-inference Public

    Multi-platform inference framework for OpenBioLLM-8B: Apple Silicon (GGUF/MLX) and GPU deployment (vLLM + FastAPI)

    Python

  4. kda-attention kda-attention Public

    Python