AI Engineer (Project based / Part Time / Freelancer)

We are seeking a skilled AI Engineer to design and build robust AI pipelines for a production-ready voice AI agent. You'll be responsible for setting up the complete architecture—from voice cloning and LLM fine-tuning to real-time STT/LLM/TTS integration—with a focus on performance, reliability, and natural interaction.

This is a hands-on, model-building role ideal for engineers who thrive on building intelligent systems from the ground up.

Responsibilities

• Design and implement end-to-end AI pipelines integrating Speech-to-Text (STT), Language Model (LLM) reasoning, and Text-to-Speech (TTS).
• Fine-tune lightweight open-source LLMs (e.g., Phi-2, Mistral, TinyLLaMA) on domain-specific conversational data.
• Clone realistic voice models using open-source TTS libraries (e.g., Coqui, Bark, TorToiSe).
• Optimize inference latency for real-time voice interaction.
• Preprocess and curate audio/text datasets for model training.
• Integrate AI components into the backend of a voice agent system.
• Set up training and deployment infrastructure using cloud GPU platforms (e.g., AWS, Lambda Labs, GCP).
• Build monitoring, logging, and testing tools for model behavior and drift.
• Collaborate with product and external telephony teams to align AI output with business goals.

Salary & Benefits

• Competitive salary and fringe benefits
• Paid Time off
• Leaves encashment
• EOBI
• Fuel Card
• Professional Development
• Career Advancement
• Team Building Activities
• Innovative Work Environment
• Work-Life Balance
• Company trips
• Wellness Programs
• Performance based promotion

Location & Work Hours

• Location: Blue Area Islamabad
• Office Hours: 05:00 PM to 02:00 AM (Monday to Friday)
• Job Types: Part-time, Temporary, Contract
• Work Location: Remote

Education & Language

• Bachelor's degree (Preferred)
• Speak English fluently (Preferred)

• Strong Python development skills, with experience in ML libraries such as PyTorch and Hugging Face Transformers.
• Hands-on experience training or fine-tuning TTS and LLM models.
• Experience building STT/LLM/TTS pipelines for real-time or near real-time systems.
• Familiarity with voice synthesis tools (Coqui TTS, Bark, etc.).
• Knowledge of prompt engineering and conversational design.
• Experience deploying models on GPU-backed environments using Docker or Conda.
• Ability to work with raw datasets (audio and text) and implement efficient data pipelines.

Nice to Have

• Experience with telephony integration (e.g., Twilio, SIP, ViciDial).
• Experience with LoRA or QLoRA for parameter-efficient fine-tuning.
• Familiarity with vector search and retrieval-augmented generation (RAG).
• Knowledge of speech prosody control and expressive speech synthesis.