3. Data, Science & AI

Adjusts NFL simulation parameters to align with target statistics such as rushing yards, passing yards, and turnover rates when deviations exceed 5%.

Validates simulation outputs against NFL statistical targets across all categories with comprehensive checks.

Analyzes JSONL benchmark data to extract and report latency, throughput, and jitter metrics for performance evaluation.

Analyzes images, videos, and audio using AI for description, transcription, and content understanding.

AI-powered visual regression testing that categorizes UI changes as expected, warnings, or errors using visual diff analysis and design token context.

AI image generation and editing using multiple models and styles.

Provides senior-level data analysis, engineering, and optimization for SQL, ETL pipelines, and data quality assurance to drive business insights.

Builds AI-powered applications with Vercel AI SDK, enabling streaming, function calling, and LLM integration for chatbots and AI features.

Transcribes audio/video to text using Whisper AI with word-level timestamps for precise synchronization.

Creates novel ML/AI tooling projects to solve unsolved problems, not best practices. Identifies gaps and builds frontier solutions.

Enables rule-based configuration for Spring AI applications, facilitating structured AI decision-making.

Provides on-device semantic search for markdown documents using BM25, vector embeddings, and LLM re-ranking for efficient knowledge base queries.

Generates optimized system prompts for AI agents and LLMs using research-backed techniques, structuring prompts with role, context, instructions, and desired output.

Imports wine data from CSV, spreadsheets, or structured text files for efficient data management.

Retrieves boat race data (starting lists, results, predictions) from CSV using date, venue, and race number parameters.

Extracts wine award data from PDF documents for competition results, ratings, and booklet processing.

Clinical decision support tool for prostate cancer treatment, covering risk stratification, LNI prediction, and drug eligibility per medical guidelines and Polish NFZ programs.

Provides AI-powered API reference from PDF documentation to resolve endpoint errors and clarify response structures.

Transforms rough prompts into production-ready versions using research-backed patterns, debugging issues and creating new prompts from scratch.

Assesses quality and effectiveness of PEIL-generated prompts, providing feedback and ratings to optimize prompt engineering for LLM applications.

Downloads and generates plots from Weights & Biases runs, visualizing metric history like loss and accuracy for machine learning experiments.

Manages AI agent context for TiDB projects using KV-cache optimization, append-only context, and memory strategies to prevent pattern lock-in.

Enables LLM fine-tuning via Tinker API, supporting supervised learning, LoRA, and vision-language models.

Audits and optimizes websites for visibility in AI search engines like ChatGPT and Google AI Overviews through GEO, structured data, and llms.txt configuration.