ABU BAKAR
[email protected] | +92-331-844-7499 | https://github.com/Abubakar12awan
Objective
AI Engineer with 3.5+ years’ experience in LLM fine-tuning, Vision Transformers and Generative Models.
Architected 20K-req/s ML pipelines on AWS SageMaker & EKS, slashing labeling costs by 80% and reducing infra
spend by 35%. Seeking to drive production-grade AI solutions and mentor cross-functional teams.
Education
● National University of Sciences and Technology (NUST) SEECS
Bachelor of Electrical/Computer Engineering
Object Oriented Programming || Machine Learning || Computer Vision || Digital Logic Design
Computer Communication and Systems || Robotics || Fundamentals of Programming || Probability
and Statistics
● Punjab group of colleges Merit Based Full Scholarship
FSC Pre-Engineering
Skills
Languages : Python | JavaScript | C++ | C | Html 5 | CSS | C#
Technologies : PyTorch | OpenCV | TensorFlow | Keras | Aws S3 || AWS EC2 ||
NumPy | Pandas | Media Pipe | Linux | Networking |
FastApi | WebSockets | MongoDB | HuggingFace |
MySQL | Docker | LLM Fine Tuning(Lora)
Generative Ai tools : APIs of Open Ai| MCP Server| Agentic models | Anthropic | Dalle
|Gemini | Eleven Labs | Llama model || Langchain | LangGraph
Version Control : Git | Github
Operating System: Linux | Windows
Security : Secure coding practices, Web and Mobile Application Security
Professional Experience
● TekStream (Private) Limited (Islamabad)
Lead Machine Learning Enginner ( Feb2024-present)
● Enterprise-scale Deployments: Architected and shipped highly available ML pipelines on
AWS SageMaker & EKS, serving 20K+ daily inference requests with >93% accuracy and
sub-40 ms latency.
● Optimization & Cost Reduction: Containerized critical services via Docker & Kubernetes;
slashed infrastructure spend by 35% using ONNX runtime and mixed-precision training.
● Innovative Modeling: Pioneered advanced Vision Transformers, GAN-based data
synthesizers, and Diffusion-driven augmentation—cutting manual labeling by 80%.
● RAG & Multi-Agent Systems: Led design and rollout of Retrieval-Augmented Generation
(ChromaDB + custom embeddings) and multi-agent chatbots (LangChain + OpenAI),
achieving 90% end-user satisfaction in enterprise support.
● MLOps Leadership: Established MLflow experiment tracking, GitHub Actions CI/CD, and
TorchServe model serving—reducing deployment cycles from weeks to hours.
● Team & Stakeholder Management: Mentor a diverse team of 8 engineers, conduct
quarterly AI workshops, and partner with product owners to align AI roadmaps with
business objectives.
● Octaloop Technologies (Islamabad)
Artificial Intelligence Engineer (May2021-Jan2024)
High-Precision Detection: Developed a 16-class room detection system using YOLOv11 &
Mask R-CNN; elevated accuracy from 52% to 96% on live architectural data.
Segmentation Excellence: Implemented UNet and SAM2 pipelines, achieving 97% IoU, and
selected UNet++ as the production standard for real-time floorplan analysis.
Advanced Architectures: Researched and prototyped Transformers, GANs, Diffusion
Models, and Autoencoders—culminating in a VAE that generated photorealistic samples on
Fashion-MNIST.
GenAI Chatbots & RAG: Built multi-agent conversational platforms with OpenAI APIs,
Faiss indexing, and LangChain orchestration—enabling contextual Q&A across global client
portals.
Full-Stack Integration: Engineered RESTful backends (FastAPI, Flask) for international
clients, deploying via Docker and ensuring 99.9% uptime.
● Eziline Software House Rawalpindi
Full Stack Developer
Develop websites Frontend and backend using Html,CSS,React jS and Node js.
● Open-source Contributor
Contributed to the AnimateAnyone Project on GitHub, showcasing collaboration skills and versatility.
Projects
● Segmentation of Walls and detections of rooms in Architectural Drawings using
UNET
● Detection of 16 distinct classes of rooms model using Yolo v11 and MaskRCCN
● AI Based healthcare Mobile App
● Chatbots tailor for specific use cases for enterprise needs.
● Develop Virtual Try on Clothes website using Sam,Deeplabv3+ and MediaPipe
● Trained Cycle Gan for style translation between images
● Integration of complex Apis.
● Autoencoders Project
We have developed a Variational Autoencoder that was able to generate realistic images using generative Ai. We
use the Fashion Mnist dataset for training.
● Unity Car Game
Make a car game in Unity Engine using C#. This game has functionalities where we have to steer cars from
different obstacles and also pass through some objects to gain reward. You can adjust your speed and direction
etc.
● Digital Twin of Induction Motor
We developed the digital Twin Induction Motor that replicates the Real-life motor variable in software using
Machine learning techniques.
Certifications
● AWS Certified Machine Learning – Specialty
● TensorFlow Developer Certificate
● Hugging Face Course: Transformer Models for NLP
● Microsoft Certified: Azure AI Engineer Associate
● Coursera: Deep Learning Specialization (deeplearning.ai)
Languages
English (Fluent), Urdu (Native)
Soft Skills
● Cross-Functional Collaboration: Effective liaison between data science, DevOps and product teams to
drive projects from proof-of-concept to production.
● Communication & Presentation: Translating complex AI/ML concepts into clear business value
propositions for both technical and non-technical stakeholders.
● Problem Solving & Critical Thinking: Rapidly debugging model performance issues, designing
experiments to isolate causes, and iterating toward optimal solutions.
