Vlad Filip@vlad_florin_filip446
1
Events attended
1
Submissions made
3+ years of experience
About me
Data and AI Engineer working in research for NLP applications in the legal industry. Managing evaluations of SOTA proprietary and open-source models. Developing frameworks for data distillation, curation and synthetic generation to fine-tune legal-specific LLMs. First-class graduate with a research-focused Masters in Engineering from the University of Cambridge. Worked in partnership with Penn State University to tackle automation and optimization for aerodynamic probes. Interned at Health Tech consultancy orchestrating 5 projects in drug and delivery unit. Palpable passion for learning, AI and self-development. Attended the European Conference of Artificial Intelligence 2025.
🤓 Latest Submissions
Mystery AI Inc. – AI Trading Agent
AI-Trading-Agent is a multi-agent cryptocurrency trading system that separates market judgment, execution planning, and risk acceptance into coordinated specialists orchestrated as a pipeline instead of one opaque model. That structure clarifies what each stage contributed, lets you tune prompts and policies independently, and supports rehearsal in paper mode before meaningful capital is at risk. The Python FastAPI backend runs analyst, trader, and risk manager stages end to end, pulls live Kraken context through the official CLI, and exposes REST endpoints for dashboards and integrations. Optional Redis stores pipeline history when configured. Multiple LLM backends are supported so teams can trade off latency, cost, and governance requirements. A SvelteKit frontend surfaces balances, market context, and recent runs for operators who prefer a UI over log diving. An optional ERC-8004 on-chain agent adds Sepolia identity binding, validates trade intents through an on-chain risk router before orders execute, and emits EIP-712 signed checkpoints for auditability. Strategy decisions still originate from the same Python service, preserving one reasoning source while benefiting from contract-level guardrails from analysis through settlement. The TypeScript agent loop can call the Python pipeline on every tick so discretionary rules, explanations, and checkpoints stay synchronized between operator tooling and on-chain execution.
12 Apr 2026
