3. Data, Science & AI

Provides database management utilities for MongoDB, including flexible schema design, aggregation pipelines, and indexing strategies with Spring Data integration.

Orchestrates multi-agent conversations among LLMs (Claude, Gemini, etc.) for collaborative problem-solving, code review, and multi-perspective task execution.

Builds AI agents with pai-agent-sdk, enabling agent creation, tool configuration, session management, and hierarchical agent structures for LLM applications.

Extracts detailed social media and web intelligence through browser automation and DOM analysis for data-driven insights.

Manages end-to-end ML lifecycle including experiment tracking, model registry, and deployment with LLM/GenAI tracing and MCP support.

Extracts eval assertions and generates test prompts from agent specs for LLM evaluation via promptfoo configurations.

Solves partial differential equations, inverse problems, and operator learning using Physics-Informed Neural Networks with PyTorch.

Coordinates multi-agent AI workflows, logs actions, manages sessions, and detects conflicts for seamless agent collaboration.

Synchronizes tasks across AI agents, logs actions, manages sessions, and detects conflicts in multi-agent workflows.

Filters collections by evaluating semantic text criteria to identify relevant items, leveraging natural language understanding.

Extracts structured metadata from notes using SQL SELECT queries, enabling efficient data retrieval and organization from database collections.

Extracts bibliographic references from PDFs using GROBID, returning structured notes per reference for academic data processing.

Enables rendering of mathematical equations and scientific notation using LaTeX/KaTeX within Slidev presentations.

Validates cryptocurrency trading strategies through historical data analysis using natural language descriptions and technical indicators like RSI and MACD.

Automates Zoho CRM lead synchronization to local SQLite with upserts, change detection, and full audit history.

Processes oversized documents by recursively splitting content and using sub-LLM calls within a REPL environment to bypass context limitations.

Framework that monitors session complexity and suggests optimal context compaction points at logical boundaries, enhancing LLM efficiency.

Initializes new repositories with AI agent rules and documentation for seamless AI agent collaboration setup.

Validates and repairs evidence references in dismech knowledge base, ensuring snippet text matches PubMed abstracts to prevent AI hallucinations.

Adds and validates ontology term references (HPO, CL, GO, MONDO, UBERON) in disorder YAML files for the dismech knowledge base.

Automates setup of experimental infrastructure for machine learning, generating fine-tuning and evaluation configurations in parallel.

Plans LLM fine-tuning and evaluation experiments, including training run configuration and generating experiment_summary.yaml files.

Executes full ML experiment workflow including model optimization and evaluation for all scaffolded runs, submitting jobs to SLURM.

Creates custom evaluation tasks for inspect-ai via an interactive, guided workflow.