Mid-Level AI Engineer Model Research & Optimization

About Us

Investic Studio develops AI-powered products across quantitative investment, financial technology, personalized AI experiences, and intelligent control systems.

Our Current AI Initiatives:

• AstroNeko: Personalized AI-powered astrology and fortune-telling
• AI Trading: Data-driven market analysis, forecasting, signal generation, and quantitative trading research
• AI Data Center Control: Intelligent monitoring, prediction, optimization, and control of data-center operations

About the Role

As a Mid-Level AI Engineer, your primary responsibility will be improving the accuracy, reliability, consistency, and practical usefulness of our AI models.

You will work with the Tech Lead, Software Engineers, Product Team, Data Team, and Domain Experts to understand problems, prepare datasets, establish evaluation methods, conduct experiments, and develop improved model solutions.

Note: This role focuses on model research, experimentation, evaluation, and optimization. You are not expected to independently manage production deployment or cloud infrastructure. However, your work should be reproducible, well-documented, and ready to be integrated into production by the engineering team.

Key Responsibilities

1. Model Research and Improvement

• Analyze existing AI models and identify weaknesses, failure patterns, and improvement opportunities
• Establish model baselines, benchmarks, evaluation datasets, and measurable success criteria
• Design and conduct controlled, reproducible experiments
• Perform qualitative and quantitative error analysis
• Research and compare appropriate models, algorithms, datasets, and techniques
• Improve model accuracy, reliability, consistency, latency, and computational efficiency
• Prepare, clean, label, validate, and version datasets
• Document experiment objectives, configurations, results, limitations, and recommendations
• Collaborate with Software Engineers to hand off models and workflows for production integration

2. AI-Powered Astrology (AstroNeko)

• Improve the accuracy, personalization, consistency, and explainability of AI-generated readings
• Develop and improve LLM, prompt, RAG, retrieval, and context-management workflows
• Translate astrology knowledge from domain experts into structured data, rules, knowledge bases, and model evaluation criteria
• Build datasets and benchmarks for evaluating astrology-related responses
• Reduce hallucinations, contradictions, repetitive answers, and unsupported claims
• Experiment with prompting, fine-tuning, structured generation, model routing, and hybrid rule-based approaches
• Improve personalization while maintaining appropriate privacy and responsible-AI practices

3. AI Trading

• Research and improve models for market analysis, forecasting, classification, ranking, sentiment analysis, and signal generation
• Work with financial time-series, price, volume, fundamental, news, and alternative datasets
• Develop proper training, validation, walk-forward testing, and backtesting methodologies
• Prevent common modelling errors such as look-ahead bias, data leakage, overfitting, and survivorship bias
• Evaluate models using both machine-learning metrics and trading-related metrics
• Include realistic assumptions such as transaction costs, slippage, latency, and market regimes
• Analyze model stability and performance under changing market conditions
• Collaborate with Quant and Domain Experts to improve trading hypotheses and evaluation criteria

4. AI Data Center Control

• Research models for anomaly detection, demand forecasting, capacity planning, predictive maintenance, and resource optimization
• Analyze operational data such as workload, compute usage, temperature, cooling, power consumption, and equipment status
• Develop models that help optimize performance, energy efficiency, capacity, and operational reliability
• Explore forecasting, optimization, reinforcement learning, or hybrid rule-based approaches where appropriate
• Evaluate model behavior under normal, abnormal, and simulated operating conditions
• Design safety constraints, fallback behavior, and human-in-the-loop controls
• Work with infrastructure and domain teams to understand operational limitations and control requirements
• Ensure proposed model decisions are measurable, explainable, and safe before production use

Qualifications

Required Qualifications:

• At least 2 years of professional experience in AI, machine learning, data science, NLP, quantitative research, or a related field

• Strong programming skills in Python

• Practical experience developing, evaluating, or improving AI/ML models

• Strong understanding of machine-learning fundamentals (Training/Validation/Testing, Feature Engineering, Overfitting, Data Leakage, Experimental Design)

• Experience with PyTorch, scikit-learn, Hugging Face, TensorFlow, or equivalent tools

• Experience preparing and analyzing structured or unstructured datasets

• Ability to define appropriate metrics, perform error analysis, and write clean, reusable code

• Ability to read research papers in English and communicate technical trade-offs clearly

• Strong practical experience in AT LEAST ONE of the following areas:

  • LLMs, NLP, RAG, or Generative AI
  • Time-series forecasting, quantitative modelling, or financial machine learning
  • Anomaly detection, optimization, control systems, or reinforcement learning

Preferred Qualifications:

• Experience with LLM prompting, evaluation, fine-tuning, LoRA, QLoRA, or PEFT
• Experience designing RAG systems, retrieval pipelines, reranking, or vector search
• Experience with financial time-series, quantitative research, or backtesting
• Experience with experiment-tracking tools such as MLflow or Weights & Biases
• Knowledge of model explainability, AI safety, guardrails, and responsible-AI practices

What Success Looks Like (First 3 Months)

• Understand the objectives, existing models, datasets, and limitations of an assigned AI product
• Establish a reproducible baseline and measurable evaluation criteria
• Identify important failure patterns through systematic error analysis
• Design and conduct experiments addressing those weaknesses
• Deliver a measurable improvement to at least one model or AI workflow
• Explain the trade-offs between model quality, latency, cost, safety, and implementation complexity
• Document the solution and collaborate with the engineering team on production integration