Daniel Kretschmer
Verified Expert in Engineering
AI Systems Architect and Quantitative Developer
Saint Julian's, Malta
Toptal member since December 8, 2025
Daniel is an AI systems architect and quantitative developer with 10+ years of experience building Python back ends and 5+ years of experience designing high-performance ML, LLM, and data pipelines. He creates production-grade AI systems from ingestion and feature engineering to training, evaluation, deployment, and monitoring. Daniel also has a strong background in quantitative research, ensuring rigorous logic, reproducible workflows, and reliable performance under real-world constraints.
Portfolio
Experience
- Python - 12 years
- FastAPI - 7 years
- Polars - 7 years
- PostgreSQL - 7 years
- Large Language Models (LLMs) - 4 years
- AI Pipeline - 4 years
- Multistage LLM Chains - 4 years
- AI Automation - 4 years
Preferred Environment
Python, Polars, Multistage LLM Chains, Large Language Models (LLMs), AI Automation, FastAPI, PostgreSQL, Docker, React, TypeScript, REST APIs, Node.js, JavaScript, MySQL, Back-end, Vite, Tailwind CSS, Data Scraping, Web Scraping, Beautiful Soup, Playwright, Selenium, Requests, Flask, LangChain, Pydantic, Open-source LLMs, AI Prompts, Agentic AI, AI Agents, OpenAI Assistants API, Zapier, n8n, Replit, Generative Artificial Intelligence (GenAI), SQL, Prompt Engineering, Charts, AI Chatbots, Large Data Sets, Full-stack Development, AI Voice Agents, Text-to-Speech (TTS), ElevenLabs Solutions, Natural Language Processing (NLP), OpenAI, Whisper, Data Science, Amazon Web Services (AWS), Amazon Bedrock AgentCore, Bayesian Inference & Modeling, Reinforcement Learning, Time Series Analysis, Statistics, Data Modeling, Financial Modeling, Normalization, Scientific Computing, Forecasting, Anthropic, Claude, Full-stack, AI Tools, AI Assistants, Bots, Data Management, Best Practices, Retrieval-augmented Generation (RAG), XGBoost, Data Analysis, PyTorch, Machine Learning Operations (MLOps), API Integration, Data Privacy, Identity & Access Management (IAM), Vector Databases, Personally Identifiable Information (PII), LlamaIndex, DevOps, Automation, Large Language Model Operations (LLMOps), Data Engineering, Data Extraction, Natural Language Toolkit (NLTK), Information Extraction, Data Labeling, PySpark, Algorithmic Trading, Pine Script, TradingView, Trading, Predictive Analytics, Data Analytics, System Modeling, Natural Language Queries, Natural Language Understanding (NLU), RAG Architecture, Apache Parquet, Pandas, Parquet, Prefect, Scikit-learn, Seaborn, Quantitative Analysis, Hetzner, Capital Structure & Products.Trade Finance
The most amazing...
...thing I've built is AlgMentor, which analyzes trader behavior using ML and data mining to detect and correct latent patterns to optimize performance.
Work Experience
Crypto & Blockchain Specialist
Freelancing as Crypto & Blockchain Specialist
- Exploited persistent mispricings between centralized and decentralized venues via triangular arbitrage, then extended to unique on-chain mechanics for predictive and exploitative edge.
- Integrated five exchanges (3 CEX, 2 DEX) in a live-deployed triangular arbitrage bot.
- Modeled execution risk with exchange-specific fee structures, API latency distributions, order book depth, and slippage curves.
- Built my own raw mempool parsers for multiple chains in JavaScript from scratch.
- Designed and deployed sniping bots, sandwich bots, and whale-alert systems tied to large pending transactions.
- Used mempool activity not just for direct MEV capture but also as a predictive signal to inform directional and hedging trades.
- Captured sub-1% triangular arbitrage spreads and multi-figure mempool MEV trades in early live environments.
- Demonstrated a repeatable process for turning raw blockchain data into actionable trading signals.
- Contributed to the entire system, which was self-researched, self-funded, and custom-built at a time when off-the-shelf bots didn’t exist.
Quantitative Researcher
Freelancing for Quantitative Research
- Tested whether institutional-grade microstructure signals (OBI, OFI, VP slope, delta, absorption, spoofing detection) retain predictive power into the 10-30 second horizon where retail-grade execution remains feasible.
- Built dozens of custom features beyond basic imbalances, including flicker detection to neutralise spoofing and engineered half-life decays to capture predictive decay curves.
- Applied regime filtering to isolate environments where signal autocorrelation is high (high vol/liquidity breaks).
- Pre-filtered features using correlation, AUC, decile lift, Granger tests, and PCA to eliminate redundancy before feeding into XGBoost with genetic and Bayesian optimisation of hyperparameters.
- Empirically derived spread, slippage, and latency distributions from my own tick-by-tick data.
- Proved that statistically significant alpha persists at 10–30 seconds even after realistic frictions. The framework demonstrates how institutional microstructure edges can be translated into deployable retail-scale bots.
AI Automation Specialist
AI Integration and Automation Freelancing
- Integrated six AI modules into a single local-first pipeline for creators that runs entirely on 8GB VRAM, free of restrictions, with modular API swaps, delivering fast, automated, pro-quality short-form videos from any transcript or long-form input.
- Built software for algorithmically analyzing latent and subconscious behavioral patterns from trade logs using AI and ML to enable traders to fix them.
- Designed, built, and validated automated trading systems using order flow, gamma exposure, and anomaly detection signals.
Lead Developer
An International Bank
- Managed a large corpus of data within the ML context.
- Investigated, reviewed, and improved upon a large legacy codebase.
- Evaluated different model performances and built reports for stakeholder review.
Experience
Fully-automated AI Content Pipeline
• A six-module pipeline for automated short-form creation
• A scalable, unrestricted creator tool for personal and commercial use with censorship-free operation
• Consumer-grade accessibility with pro output
• End-to-end automation from transcript to video
• Llama-3.1 instructions with auto-validation and retries
• ComfyUI workflows for visuals and video assets
• ChatterboxTTS zero-shot cloning superior to ElevenLabs
• WhisperX microsecond captions and MusicGen adaptive music
• Direct YouTube transcript pull with highlight extraction
• Complete short-form packages: visuals, captions, voice, and music
• Copyright-free, offline or hybrid deployment
• Pro-grade quality on consumer hardware
It demonstrates a creator-scale application of advanced AI pipelines.
FIREForge
https://fireforge.meDue to advanced back-end engineering, results are returned instantly, eliminating waiting times for backtests to conclude, and enabling exploratory use of the product.
It further supports regime filtering and leverages forward testing with detailed metrics to ensure that results align with expectations.
AlgMentor
Beyond that, it includes a heuristic model that tells users, based on the results of their live trading session, whether they should continue or not.
Skills
Libraries/APIs
NumPy, TensorFlow, REST APIs, Node.js, Beautiful Soup, Playwright, Requests, XGBoost, Natural Language Toolkit (NLTK), Pandas, Scikit-learn, React, Pydantic, OpenAI Assistants API, PyTorch, PySpark
Tools
Whisper, Claude, Seaborn, AI Prompts, Zapier, n8n, Prefect
Languages
Python, JavaScript, SQL, Pine Script, TypeScript, Solidity
Frameworks
Selenium, Flask, LangGraph, Tailwind CSS, LlamaIndex
Paradigms
Best Practices, Automation, Microservices, DevOps, Quantitative Research
Storage
Data Pipelines, MySQL, Apache Parquet, PostgreSQL, Databases
Platforms
Docker, Replit, Amazon Web Services (AWS), Blockchain
Other
Multistage LLM Chains, Large Language Models (LLMs), AI Automation, AI Pipeline, Artificial Intelligence (AI), Machine Learning, Cursor AI, Back-end, Data Scraping, Web Scraping, Quantitative Analysis, Agentic AI, AI Agents, Generative Artificial Intelligence (GenAI), Prompt Engineering, Stock Trading, Charts, AI Chatbots, Large Data Sets, Stock Market, Full-stack Development, AI Voice Agents, Speech-to-Text (STT), Text-to-Speech (TTS), Natural Language Processing (NLP), OpenAI, Data Science, Bayesian Inference & Modeling, Reinforcement Learning, Time Series Analysis, Statistics, Data Modeling, Financial Modeling, Performance Optimization, Scientific Computing, Financial Forecasting, Forecasting, Mathematical Finance, Full-stack, AI Tools, AI Assistants, Bots, Data Management, IT Management, Retrieval-augmented Generation (RAG), Data Analysis, Machine Learning Operations (MLOps), API Integration, Data Privacy, Identity & Access Management (IAM), Vector Databases, Personally Identifiable Information (PII), Large Language Model Operations (LLMOps), Data Engineering, Data Extraction, Algorithmic Trading, TradingView, Trading, Predictive Analytics, Data Analytics, System Modeling, Natural Language Queries, Natural Language Understanding (NLU), RAG Architecture, Parquet, Hetzner, Capital Structure & Products.Trade Finance, AI Systems, Multi-agent Orchestration, Multi-agent Systems, Agentic AI Systems, Polars, FastAPI, Data Mining, SaaS, Web UX, APIs, Computer Vision, Supabase, Vite, LangChain, Open-source LLMs, ElevenLabs Solutions, Normalization, Anthropic, RAG Pipelines, Information Extraction, Data Labeling, Multimedia, Finance, Research, Quantitative Finance, Quantitative Modeling, Quantitative Risk Analysis, Decentralized Finance (DeFi), Arbitrage, Statistical Arbitrage, Amazon Bedrock AgentCore
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring