Muhammad Asad

Software Engineer specializing in AI Engineering and Agentic Workflows

Based in: Lahore, PakistanCurrently: Community Member, HAMiExperience: 2 years

Hire me

Software Engineer with expertise in building AI assistants, RAG applications, and agent workflows using LangChain and LangGraph. He has a strong background in full-stack development and open-source contributions to Kubernetes and cloud infrastructure projects.

Experience

Community Member

HAMivolunteerSep 2026 – Present

Summary

Contribute to HAMi's open source GPU infrastructure and work with maintainers to improve how Kubernetes manages and shares AI hardware.

What I did

  • Work on Kubernetes scheduling and resource accounting for shared GPU workloads
  • Contribute across NVIDIA MIG, AWS Neuron, Cambricon MLU, MetaX, and other accelerator integrations
  • Improve system reliability through configuration validation, edge-case testing, Helm checks, and maintainer-reviewed changes

Results

  • Had multiple PRs merged into the project across scheduler internals, device resource management, and workload reliability.
KubernetesGPUNVIDIA MIGAWS NeuronCambricon MLUMetaXHelm

MERN Stack Developer

Dev WeekendsLahore, PakistanworkAug 2024 – Present

Summary

Deliver practical software engineering training and help learners turn technical concepts into working applications.

What I did

  • Lead hands-on sessions covering web development, backend systems, mobile applications, and software engineering practices
  • Guide learners through real projects involving application design, API integration, debugging, and deployment
  • Review code, explain technical problems clearly, and help learners improve their problem-solving approach
MERN StackWeb DevelopmentBackend SystemsMobile Applications

AI Reasoning Engineer

TuringPalo Alto, United States (Remote)workAug 2025 – Feb 2026

Summary

Trained and evaluated AI reasoning models using competitive programming, high-performance C++, and reproducible Docker environments.

What I did

  • Designed complex algorithmic problems to test logical reasoning, optimization, and performance under time and memory limits
  • Analyzed model outputs, identified reasoning gaps, and created targeted C++ tasks with efficient reference solutions
  • Built Docker-based evaluation workflows and worked with AI research teams to improve evaluation methods, data quality, and model reliabilityodel's logical reasoning, optimization behavior, and overall problem-solving depth.
  • Created complex C++ reasoning problems, reviewed model-generated solutions, and identified reasoning failures.
  • Validated correctness of model outputs against edge cases and time/space constraints.
  • Worked on LLM evaluation and improving training tasks and data to enhance reasoning quality and accuracy.
C++DockerAI Reasoning Models

Skills

Technical

JavascriptJavascript
ReactJSReactJS
GitGit
RAG
CSSCSS
LangChain
HTMLHTML
tailwind CSStailwind CSS
Lang Graph
BootstrapBootstrap
Node.jsNode.js
Hugging Face
Express.jsExpress.js
MongoDBMongoDB
FastAPIFastAPI
Postman
Stripe
React ReduxReact Redux
ChromaDB
FirebaseFirebase
AWS
Microsoft Azure
PayPal
Gradio
LLM Trainer
FAISS
RLHF
SQLiteSQLite
GitHubGitHub
C++C++
Semantic retrieval
Vector database
TensorFlow LiteTensorFlow Lite
Pandas
NumPy
Model evaluation
LoRA
PEFT

Areas of expertise

Problem Solving
Debugging Code
Analytical Solutions
Collaboration

Projects

HAMi | Kubernetes GPU Resource Management

Contributoropen_source
HAMi is an open source Kubernetes project that helps teams share and manage GPUs and other AI accelerator resources across workloads. Contributed to GPU scheduling, resource accounting, and dynamic device management. Worked across NVIDIA MIG, AWS Neuron, Cambricon MLU, MetaX, and other accelerator integrations. Improved configuration validation, automated testing, Helm checks, and system reliability.
KubernetesGoGPU SchedulingHelmDockerGitHub

Inroad | AI-Powered Job Search Platform

Full Stack Developerprofessional
Inroad is a job search platform that matches users with suitable companies, researches each company, and creates personalized outreach content. Redesigned profile, contact, and settings pages with responsive and accessible user flows. Improved résumé upload handling by connecting typed FastAPI errors with clear frontend feedback. Worked across the Next.js and TypeScript frontend and Python and FastAPI backend.
Next.jsTypeScriptReactPythonFastAPIPostgreSQLDockerVector databaseEmbeddingsSemantic retrieval

LearnForge LMS

Full Stack Developerpersonal
LMS is a full stack online learning platform where students can discover, purchase, and complete courses while administrators manage content, users, orders, and platform activity. Built student flows for course discovery, secure payments, video learning, questions, reviews, and progress tracking. Developed an administration dashboard for managing courses, users, orders, categories, notifications, and analytics. Implemented secure authentication, Stripe payments, Redis caching, media storage, and real-time updates.
Next.jsReactTypeScriptNode.jsExpressMongoDBRedisStripeSocket.IO

MarketLane

Full Stack Developerpersonal
MarketLane is a full stack marketplace that connects customers and independent sellers through one platform for shopping, payments, order management, and communication. Built separate customer and seller experiences with secure authentication and role-based access. Developed product management, cart, checkout, order tracking, refunds, seller balances, and withdrawal workflows. Integrated Stripe, PayPal, Cash on Delivery, Cloudinary media storage, and real-time buyer-seller messaging.
ReactRedux ToolkitNode.jsExpressMongoDBSocket.IOStripePayPalCloudinary

Apache DolphinScheduler | Workflow Orchestration Platform

Contributoropen_source
Apache DolphinScheduler is an open source platform for creating, scheduling, and monitoring complex data workflows. Added timeout handling and automated tests for script-based alert execution. Improved workflow loading states and added clearer warnings for destructive actions. Removed unused alert code and improved reliability across workflow management features.
JavaVue.jsTypeScriptWorkflow SchedulingDistributed SystemsGitHub

Plant Leaf Disease Detection

personal
Built a Flutter application that integrated a TensorFlow Lite model for real-time image-based predictions of plant diseases.
TensorFlow LiteFlutterPythonPandasNumPy

Education

BS in Computer Science

University of Education LahoreComputer Science

ICS

Punjab College Kasur

Recognition

Awards

ICPC Regional Qualifier

International Collegiate Programming Contest

Qualified for the ICPC regional stage by solving complex algorithmic problems under strict time limits. Applied data structures, algorithms, logical reasoning, and performance optimization while working in a competitive team environment.

Certifications

Flutter UI BootCamp

Udemy

Next js

Coursera

Microsoft Certified: Azure Fundamentals

MicroSoft

IT Specialist - Data Analytics

Certiport - A Pearson VUE Business
Hire MuhammadReach out about a role, a contract or a conversation.For recruiters

Muhammad's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis