Custom Python LLM Creation for PositoAI
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted1 hour ago
Project Title
Machine Learning Engineer to Build and Train Custom Python LLM for PositoAI.com
Project Description
We run PositoAI.com, an AI-powered stock research and screening platform. We want to move away from commercial APIs (like OpenAI or Anthropic) and build, fine-tune, and deploy our own custom Large Language Model (LLM).
We are looking for an experienced Machine Learning / Deep Learning Engineer to build a Python-based pipeline that trains a custom open-source model to auto-generate structured, accurate financial stock summaries.
1. Scope of Work (What We Need Done)
• Model Selection & Architecture: Help us select the best open-source base model (e.g., Llama 3, Mistral, Phi-3, or a financial variant like FinGPT) to host and train locally or on our cloud infrastructure.
• Dataset Preparation: Build Python pipelines to clean, structure, and tokenize our financial data (ratios, metrics, text-based disclosures) into training datasets.
• Fine-Tuning / Training: Implement fine-tuning strategies—such as LoRA, QLoRA, or Full Parameter Fine-Tuning—using Python to teach the model to reliably interpret financial metrics and generate natural summaries.
• Deployment & Inference: Set up a highly optimized self-hosted inference pipeline (using tools like vLLM, Ollama, or Hugging Face TGI) to serve predictions efficiently with low latency.
• Guardrails & Evaluation: Implement strict Python guardrails to evaluate text generation and minimize hallucinations to ensure strict alignment with financial domain data.
2. Technical Requirements
• Core Language: Advanced, production-level Python.
• ML Frameworks: PyTorch, Hugging Face Transformers, PEFT, TRL, or Accelerate.
• Inference Optimization: Experience with vLLM, TensorRT-LLM, or DeepSpeed.
• Infrastructure: Familiarity with training models on GPU cloud platforms (AWS, RunPod, Vast.ai, or Lambda Labs).
3. What You Need to Provide in Your Proposal (Mandatory)
To be considered for this role, you must explicitly state:
1. Duration: How many weeks will it take you to deliver a trained, deployable model and a working inference pipeline?
2. Total Cost: What is your exact fixed price for this custom ML development?
3. Experience: Please detail your past experience with training or fine-tuning open-source LLMs using PyTorch and Hugging Face.
Machine Learning Engineer to Build and Train Custom Python LLM for PositoAI.com
Project Description
We run PositoAI.com, an AI-powered stock research and screening platform. We want to move away from commercial APIs (like OpenAI or Anthropic) and build, fine-tune, and deploy our own custom Large Language Model (LLM).
We are looking for an experienced Machine Learning / Deep Learning Engineer to build a Python-based pipeline that trains a custom open-source model to auto-generate structured, accurate financial stock summaries.
1. Scope of Work (What We Need Done)
• Model Selection & Architecture: Help us select the best open-source base model (e.g., Llama 3, Mistral, Phi-3, or a financial variant like FinGPT) to host and train locally or on our cloud infrastructure.
• Dataset Preparation: Build Python pipelines to clean, structure, and tokenize our financial data (ratios, metrics, text-based disclosures) into training datasets.
• Fine-Tuning / Training: Implement fine-tuning strategies—such as LoRA, QLoRA, or Full Parameter Fine-Tuning—using Python to teach the model to reliably interpret financial metrics and generate natural summaries.
• Deployment & Inference: Set up a highly optimized self-hosted inference pipeline (using tools like vLLM, Ollama, or Hugging Face TGI) to serve predictions efficiently with low latency.
• Guardrails & Evaluation: Implement strict Python guardrails to evaluate text generation and minimize hallucinations to ensure strict alignment with financial domain data.
2. Technical Requirements
• Core Language: Advanced, production-level Python.
• ML Frameworks: PyTorch, Hugging Face Transformers, PEFT, TRL, or Accelerate.
• Inference Optimization: Experience with vLLM, TensorRT-LLM, or DeepSpeed.
• Infrastructure: Familiarity with training models on GPU cloud platforms (AWS, RunPod, Vast.ai, or Lambda Labs).
3. What You Need to Provide in Your Proposal (Mandatory)
To be considered for this role, you must explicitly state:
1. Duration: How many weeks will it take you to deliver a trained, deployable model and a working inference pipeline?
2. Total Cost: What is your exact fixed price for this custom ML development?
3. Experience: Please detail your past experience with training or fine-tuning open-source LLMs using PyTorch and Hugging Face.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.