AI Algorithmic Trader -- Interview Today, Start Tomorrow!

Job Description

Job Description

AI Algorithmic Trader — MemeLabs, Inc.

MemeLabs, Inc., a new startup division of MemeHouse LA, is hiring a systematic trader-researcher to answer one question: can an AI learn to trade and improve itself, or does it only look like it is learning? You will design the strategies, the tests and the learning loop, on the world's largest DEXs.

PLEASE ONLY APPLY IF YOU ARE ABLE TO START RIGHT AWAY

This role starts immediately. The 45- and 90-day milestones run from your first day, so we can only consider candidates who are available to begin now. If you need a notice period or a later start date, please do not apply for this opening.

Job details

Company

MemeLabs, Inc. (a startup division of MemeHouse LA)

Location

Beverly Hills, CA — in person and remote

Employment type

Full-time, part-time, W-2 employee, or contract

Annual base salary

$120,000–$280,000

Additional compensation

$5,000 performance-based signing incentive, earned through two initial-delivery milestones targeted for the first 90 days

About us

MemeHouse LA is launching a startup division, MemeLabs, Inc., where we build AI software that turns information into useful decisions and reliable workflows. Our focus is practical: secure AI deployments, connected systems, and applications that deliver measurable value.

One of those applications is a systematic crypto trading program on the world's largest decentralized exchanges, across BTC, ETH, SOL and other major markets. An AI agent writes the backtests, bots, analysis and hourly market reads; it keeps a memory of what it tried and grades its own past predictions.

So far the evidence says the AI is not yet learning, which is why this role exists:

• A grid fitted to our in-house AI market model's levels made +0.43%/day on the data it was tuned on and −0.25%/day on the next eight weeks.
• The model's trade setups, tested on 1,497 trades, matched random entries with the same stops and targets almost to the basis point.

We want someone who finds these results interesting, not discouraging.

About the role

As our AI Algorithmic Trader, you are the human in the loop between an AI that can build a trading bot in an hour and a market that punishes anything untested. You will design systematic strategies, build the evaluation that decides which ones are real, and engineer the loop that lets the AI learn from every result.

This is a hands-on role with ownership of meaningful deliverables. You will explain trade-offs to non-traders, document decisions, and help establish repeatable practices as the team grows.

What you will do

• Design strategies: Create systematic strategies with clear entry, exit, sizing and kill rules, and turn each into a testable hypothesis.
• Build the self-improvement loop: Decide what the AI remembers, how it grades its own calls, which changes it may make alone and which need a human; then prove with numbers whether it improves.
• Evaluate quality: Build repeatable backtests with realistic fills, fees and slippage, walk-forward splits, held-out data, and a control for every result (random entries, random levels, a plain grid).
• Grade the AI market model: Score its levels, setups and stated odds against simple baselines, and decide what, if anything, should drive bots.
• Run paper and live bots: Move survivors to paper beside a control, then small live size; monitor, scale or retire them.
• Operate dependable systems: Exchange-side stops, software loss limits, a 90%-of-equity deploy cap, observability, versioning and rollback for every bot.
• Protect sensitive information: Least-privilege keys, secrets management, dry-run-first transfer scripts, and safeguards against prompt injection into the AI's tools.
• Improve the function: Review code, document decisions in shared memory, and help mentor additional researchers as the team expands.

Strategies you'll build and test

Family

Examples we run or have tested

Grid and ladder

Static and dynamic grids, re-centring ping-pong ladders, staggered take-profits across many small orders

Mean reversion

Buying support and fading resistance from model-picked levels, VWAP and range fades

Breakout and momentum

Close-above / close-below triggers, trailing take-profits, trend-gated entries

Accumulation

DCA and TWAP drips with price gates and breakeven stops

Signal-driven

On-chain flows, funding, open interest, liquidation clusters, stablecoin dominance, TradingView alerts

Regime switching

Hourly on/off switchboards, volatility and chop filters, scheduled-event kill switches

The interesting work is not any single strategy. It is building the process that tells the AI which ideas work, why, and what to try next.

What we are looking for

• Systematic strategy experience: 3+ years designing, building and running your own systematic trading strategies, with live results you can walk us through, across several types (grid, mean reversion, momentum, signal-driven), including ones that failed and why.
• Software engineering ability: Strong Python; you can read a 600-line bot and spot the bug in its order logic, and you write maintainable code, APIs and automated tests.
• Statistics you actually use: Out-of-sample testing, noise vs edge, power and sample size, overfitting and multiple comparisons.
• Exchange and execution know-how: Order types, post-only and reduce-only, trigger stops, rate limits, partial fills.
• Applied AI experience: Working fluency with AI coding agents and LLM tooling, plus the habit of verifying their output instead of trusting it.
• Backend fundamentals: Comfort with SQL, Git, Linux, Docker, API authentication and scheduled jobs.
• Security awareness: Sound handling of keys, wallets, permissions and model limitations.
• Ownership and communication: You scope work, raise blockers early, document decisions, and kill your own strategies when the data says so.
• Demonstrated capability: Relevant professional experience, substantial project work, or equivalent evidence. A specific degree is not required.

Helpful additional experience

• Self-improving agents: Memory, reflection, graded predictions, prompt or workflow evolution, acceptance tests.
• Market microstructure: Order-book depth, queue position, liquidation cascades, funding.
• On-chain DEX trading: Order-book DEXs and their SDKs, wallets, transfers and withdrawals.
• Market data: On-chain and derivatives data (Glassnode, CoinGlass) and a view on what in them is signal.
• Forward testing: A paper-trading or shadow-trading harness you have built.
• Team contribution: Open-source work, technical writing, or mentoring.

AI models, tools and infrastructure

We expect hands-on familiarity with many of the tools below and real depth in a few. Tell us which ones you have used in production and for what.

Area

Tools and technologies

Frontier LLMs and assistants

Claude (Anthropic), ChatGPT and GPT models (OpenAI), Grok (xAI), Gemini (Google), Mistral, DeepSeek

Open-weight models

Llama (Meta), Hermes (Nous Research), Qwen, Mistral and Mixtral, DeepSeek, Gemma, Phi

AI coding agents

Codex (OpenAI), Claude Code, Cursor, GitHub Copilot, Aider

Agent frameworks and protocols

Claude Agent SDK, OpenAI Agents SDK, Model Context Protocol (MCP), LangChain and LangGraph, LlamaIndex, CrewAI, AutoGen

Agent memory and retrieval

Embeddings, RAG, vector stores (pgvector, Qdrant, Chroma, Pinecone), structured memory files, reflection loops

Training and fine-tuning

PyTorch, Hugging Face Transformers, TRL and PEFT, LoRA and QLoRA, Unsloth, Axolotl, RLHF and DPO

Reinforcement learning for trading

Gymnasium, Stable-Baselines3, Ray RLlib, FinRL

Model serving and inference

vLLM, SGLang, llama.cpp, Ollama, Hugging Face TGI, TensorRT-LLM; quantization (GGUF, AWQ, GPTQ)

Evaluation and observability

Weights & Biases, MLflow, LangSmith, Langfuse, promptfoo; graded-prediction and acceptance-test harnesses

Quant and backtesting

pandas, NumPy, Polars, DuckDB, vectorbt, backtrader, freqtrade, Hummingbot, QuantConnect, TradingView and Pine Script

Exchange and on-chain

DEX and CEX REST/WebSocket APIs, ccxt, web3.py, ethers.js, wallet and key management, bridges

Market data

Glassnode, CoinGlass, exchange order-book and trade feeds, funding and open-interest data

Hardware

NVIDIA GPUs (RTX 4090/5090, A100, H100) with CUDA; Apple Silicon (M-series, MLX) for local models; always-on Linux or macOS servers; low-latency VPS near exchange endpoints; hardware wallets

Cloud and operations

AWS, GCP, Lambda or RunPod GPU cloud, Docker, Kubernetes, Linux, Git, CI/CD, cron and launchd schedulers, ngrok or Cloudflare tunnels, secrets managers (1Password, Vault)

How we work

1. Evidence decides. A result counts only out-of-sample, after fees, against a control.
2. Tune on one period, judge on the next. In-sample winners are a hypothesis, not a strategy.
3. Paper before live, small before large. A new bot earns size over weeks, not hours.
4. The AI builds; a human gates money. Agents research, code and test. Launching, sizing and moving funds stay with people.
5. Ask the exchange, don't model it. Verify every stop is live and every fill is real.
6. Write it down. Every test, result and mistake goes into shared memory so it is never repeated.

Your first 90 days

• First 30 days: Learn the strategy library, bots, AI memory files and risk controls; audit which backtest and live numbers you trust and why; set up your environment and evaluation approach.
• Target by day 45 (milestone one): A shared test harness with controls built in (random entries, random levels, plain grid), plus a scorecard that tracks whether the AI's ideas and market calls are improving, with reproducible setup and first results.
• Target by day 90 (milestone two): At least one strategy that beats its control on unseen data and in paper trading, running live at small size with monitoring, kill switches and a launch-and-rollback runbook, and a written first answer on whether the AI is improving.

Compensation and $5,000 signing incentive

The annual base salary range is $120,000–$280,000. The starting offer will reflect the responsibilities of the position, demonstrated skills, relevant experience, and internal pay equity; the $5,000 incentive is separate from base salary.

We offer a $5,000 gross performance-based signing incentive, structured as a new-hire milestone bonus rather than an upfront payment:

• Milestone one: $2,500. Earned on completion of the agreed test harness, AI scorecard and initial results, targeted for day 45.
• Milestone two: $2,500. Earned on completion of the agreed validated strategy, live monitoring, runbook and written AI-improvement assessment, targeted for day 90.

Written milestone requirements and objective acceptance criteria will be provided before employment begins. Each installment is earned when its criteria are satisfied and will be paid on the next regular payday, or earlier if required by applicable law, less required tax withholdings.

The 45- and 90-day dates are delivery targets, not minimum-service or automatic-forfeiture dates. Once earned, an installment is not conditioned on continued employment through its payment date and does not have to be repaid if employment later ends. Milestones measure delivery of the agreed work, not trading profit.

Benefits

• Health coverage: Employer-sponsored health insurance, subject to plan eligibility and terms.
• Retirement benefits: 401(k) participation, subject to plan eligibility and terms.
• Time off: Unlimited paid time off under the company's written policy, plus applicable paid sick leave and other legally protected leave.

Why join us?

You will own a live experiment in whether AI can learn to trade, and shape how it is built and run. We value clear thinking, candid communication, responsible handling of capital and data, and strategies that hold up beyond the backtest.

Equal opportunity and accommodations

MemeLabs, Inc. is an equal opportunity employer. We consider qualified applicants without discrimination based on any characteristic protected by applicable federal, state, or local law and provide reasonable accommodations during the application process and employment.

How to apply

Submit your resume, a GitHub profile or portfolio if available, and one page on a systematic strategy you built and later killed: what it was, what the backtest said, and what made you stop it. Applications to: View email address on us.fitly.work. Please do not submit confidential information or proprietary code belonging to a current or former employer or client.