At ACRES I operate at two altitudes simultaneously. Down in infrastructure: perpetual trading bots executing across six DEX platforms (Lighter, Apex, Extended, Nado, Variational), each with their own quirks, rate limits, and failure modes. The bots don't just trade. They manage risk, handle outages, and adapt to market conditions.
Up in the intelligence layer: I fine-tuned Qwen 3.6 and Gemma 4 with Unsloth + LoRA on our internal codebases and trading logs, then ran inference locally across llama.cpp (with llama-server), Ollama, LM Studio, and MLX on Apple Silicon, depending on the deployment target. Everything performance-critical or sensitive runs on weights we own, bypassing API rate limits and keeping proprietary signals out of third-party providers.
On top of that sits the agent layer: Hermes agents handle orchestration, function calling, and structured outputs that drive code review and trade execution. I built the eval harness too: curated datasets, agent benchmarks, DPO loops to keep the models honest. Without evals, "AI infrastructure" is just vibes; with them, you actually know what your stack is doing.
This is the convergence. Blockchain proved code can enforce rules. AI proves code can make judgment calls, when you train and serve the models yourself instead of renting them. I'm building where those two forces meet.
6dex integrations
24/7autonomous bots
2models fine-tuned · Qwen / Gemma
llama.cppself-hosted inference
Chronologic AI Pipeline
Internal AI infrastructure pairing self-hosted Qwen + Gemma with Hermes-orchestrated agents, automating code reviews, documentation, and deployment workflows across the engineering team.
visit the Chronologic pipeline site →
Internal LLM & Agent Stack
Fine-tuned Qwen 3.6 + Gemma 4 with Unsloth/LoRA on internal data; served locally via llama.cpp, Ollama, LM Studio, and MLX on Apple Silicon. Hermes agents wired into engineering and trading workflows. Custom eval harness with DPO loops for continuous improvement.
Automated Perpetual Trading System
Cross-DEX trading bots with execution strategies, risk management, and automated airdrop farming.
Perp Trading
Qwen 3.6 · Gemma 4
Unsloth · LoRA
llama.cpp · llama-server
Ollama · LM Studio · MLX
Hermes Agents
DPO · Evals