Posted Oct 21, 2025 118 views
We are seeking a skilled AI Model Optimization & Fine-Tuning Engineer to join our team, specializing in fully offline, on-device deployment of large AI models. The ideal candidate will have practical experience in fine-tuning, distilling, and quantizing large base models tailored for mobile platforms such as Android and iOS. This role focuses on optimizing models for efficient mobile inference, addressing challenges related to battery consumption, latency, and memory constraints. The position requires expertise in model compression, dataset preparation, and benchmarking on real devices to ensure high performance and usability.
Key Responsibilities
- Oversee the complete optimization pipeline, including data preparation, fine-tuning, distillation, quantization, mobile packaging, and benchmarking of AI models.
- Implement Post-Training Quantization (PTQ) and Quantization-Aware Training (QAT) techniques, utilizing methods such as AWQ, GPTQ, SmoothQuant, and RPTQ to deploy large language models (LLMs) and multimodal models on devices with limited memory and thermal budgets.
- Prepare and format datasets for fine-tuning, covering tokenization, tagging, deduplication, and version control processes.
- Optimize models to maximize battery efficiency, reduce latency, and minimize RAM usage on mobile devices.
- Conduct benchmarking and debugging of inference performance using tools like Perfetto, Battery Historian, and Instruments.
- Collaborate closely with application development teams to integrate optimized models seamlessly into mobile applications.
Required Qualifications
- Demonstrated experience with quantization and distillation techniques, including PTQ, QAT, and 4-bit/8-bit schemes such as INT4, INT8, FP4, and NF4.
- Proficiency in distillation methods including teacher–student models, logit matching, and feature distillation.
- Expertise in fine-tuning using LoRA, QLoRA, DoRA, AdaLoRA, and instruction-tuning pipelines built on PyTorch and Hugging Face frameworks.
- Strong skills in dataset formatting (e.g., JSONL, multi-turn dialogs), tokenization (SentencePiece, BPE), deduplication, stratified sampling, and evaluation set creation.
- Hands-on experience deploying models on at least two runtimes such as llama.cpp/GGUF, MLC LLM, ExecuTorch, ONNX Runtime Mobile, TensorFlow Lite, or Core ML.
- Knowledge of hardware acceleration technologies including Metal (iOS), NNAPI (Android), GPU/Vulkan, Qualcomm DSP/NPU, and XNNPACK.
- Proven track record of real-world deployment of fully offline models on mobile devices, with documented performance metrics covering tokens per second, RAM usage, and device specifications.
- Proficiency in performance benchmarking and profiling using tools like Perfetto, systrace, Battery Historian, adb stats (Android), Xcode Instruments, and Energy Log (iOS).
- Strong experience with PyTorch and Hugging Face, coupled with the ability to clearly document and explain optimization trade-offs.
Preferred Qualifications and Benefits
- Contributions to open-source projects related to LLM quantization or edge AI frameworks.
- Experience deploying models from Qwen, LLaMA, Gemma, or Mistral families on mobile platforms.
- Background working with multilingual or low-resource datasets such as Urdu, Arabic, or Hindi, including expertise in tokenization, script handling, and fine-tuning.
- Familiarity with on-device multimodal integration, including ASR (automatic speech recognition), TTS (text-to-speech), and VAD (voice activity detection).
Applicants are requested to submit a brief case study demonstrating a fine-tuned model, detailing the dataset, methodology, and results. Additionally, a case study on a quantized or distilled model for mobile deployment should be provided, specifying the framework, bit-depth, device, and performance metrics. Sharing links to GitHub repositories, research papers, or APK/TestFlight builds is highly encouraged.
This is a full-time, in-person role offering a competitive monthly salary range of Rs250,000 to Rs400,000.
Skills
Job Details
Sublime Wireless
What is your Competitive Advantage?
Zohaib Ali
Marketing Manager
Digital Online Devices
Ali Aziz
CEO
Shopistan
Sehrish Azhar
Motivational Speaker & Blogger
Living Upbeat
Muzammil Baig
Digital Marketing Manager
IT Services Group
Aqib Irshad
Head of HR, Training & OD
FINCA
Ayesha Shaukat
Social Media Marketing Manager
Ivoke
Usman Habib
Recruitment & Training Executive
ACS Synergy (Pvt) Ltd.
Mona Sharif
Senior Team Lead
Konext PVT. LTD
Zubair Anjum
Business Excellence Specialist
Nestle Pakistan
Babar Ahmed Khan
Territory Sales Manager
Engro Foods
