AI Training & Fine-Tuning Specialist

About Us & Our Mission 

At our core, we build specialized, high-performing AI solutions that solve complex enterprise automation challenges. Rather than relying solely on off-the-shelf foundation models, we focus on domain-adapted, fine-tuned models that deliver domain accuracy, reduced latency, and strict data privacy. We foster an engineering culture grounded in craftsmanship, technical rigor, and empirical evaluation—where real deliverables and proof-of-work take precedence over legacy credentials or corporate fluff.

Role Summary 

We are seeking an AI Training & Fine-Tuning Specialist to take ownership of our end-to-end model adaptation pipeline. In this role, you will curate high-quality datasets, execute parameter-efficient fine-tuning (PEFT, LoRA/QLoRA), conduct alignment techniques (RLHF, DPO, ORPO), and build rigorous evaluation benchmarks. You will work closely with domain experts and software engineers to ensure our custom models achieve domain-specific precision, low hallucination rates, and optimal performance on resource-constrained deployments.

Key Responsibilities 

Data Curation & Quality Engineering: Design, clean, annotate, and augment high-quality instruction-tuning and preference datasets. Implement automated QA filtering to eliminate noise and bias. Model Training & Fine-Tuning: Execute fine-tuning pipelines for open-weights Large Language Models (e.g., Llama, Mistral, Qwen) using techniques like LoRA, QLoRA, Full Parameter Tuning, and DPO/RLHF. Evaluation & Benchmarking: Establish automated evaluation suites (e.g., MT-Bench, lm-evaluation-harness, custom domain metrics) to measure model accuracy, latency, safety, and hallucination rates. Optimization & Inference Strategy: Quantize and optimize trained models (e.g., GGUF, AWQ, TensorRT-LLM, vLLM) for high-throughput, low-latency production serving. Cross-Functional Collaboration: Partner with backend engineers to integrate fine-tuned models into live API pipelines and feedback loops.

Qualifications & Skills 

Must-Have Requirements (Non-Negotiable) Hands-On Framework Experience: 2+ years of hands-on experience training, fine-tuning, or evaluating LLMs using PyTorch, Hugging Face (Transformers, TRL, PEFT), and Unsloth or Axolotl. Fine-Tuning Techniques: Demonstrated expertise in parameter-efficient techniques (LoRA, QLoRA) and preference optimization (DPO, IPO, or RLHF). Data Quality & Curation: Experience designing instruction datasets, synthetic data generation pipelines, or filtering large unstructured text corpora for model training. Evaluation & Metrics: Strong understanding of LLM evaluation frameworks, automated evaluation techniques (LLM-as-a-Judge), and classical NLP metrics (ROUGE, BLEU, perplexity). Python & Tooling: Advanced Python proficiency with experience using data processing libraries (Pandas, Polars, Datasets) and Git version control. Nice-to-Have Bonus Skills Experience with distributed training frameworks (DeepSpeed, PyTorch FSDP, Ray). Familiarity with CUDA optimization, kernel tuning, or quantization frameworks (vLLM, TensorRT-LLM). Contributions to open-source model releases, dataset publications, or AI research.

Multistage Hiring Process (Proof-of-Work Evaluation)

 To ensure fair evaluation and eliminate resume-screening bias, we use a streamlined, multi-stage hiring process: Initial Review: Brief assessment of experience and public projects (GitHub, Hugging Face, or portfolio). Intermediate Practical Assessment: A short, paid work-sample task (e.g., analyzing a dataset filtering challenge or configuring a LoRA fine-tuning script). Technical Interview & Deep Dive: A 45-minute discussion on model architecture trade-offs, data quality strategies, and evaluation methodology. Offer & Onboarding: Clear offer presentation with transparent compensation and stock allocation.

Benefits & Compensation Transparent Pay:

 Competitive base salary (120k–165k USD) aligned with market benchmarking data. Workplace Flexibility: 100% remote work with flexible asynchronous hours. Professional Growth: $2,500 annual budget for compute credits, conferences, research papers, and training. Health & Wellness: Comprehensive health, dental, and vision insurance coverage (or stipend equivalent for global hires). Equipment Allowance: Top-tier workstation setup provided (MacBook Pro / Linux workstation + dedicated GPU cloud compute credits).

How to Apply To apply, please submit your resume or a link to your GitHub / Hugging Face profile. We welcome applications from self-taught engineers, research professionals, and non-traditional backgrounds who can demonstrate practical fine-tuning and data curation competency.