Head of Applied AI & Trading Systems (Iron Man Track)

Job description

## **The Gist** Deeter Analytics is a private investment firm managing approximately **$500M** across public markets, alternatives, and private investments. We are small, profitable, and aggressive. We don’t have committees. We have capital, judgment, and a need for speed. We are hiring a **technical builder** to sit directly beside the CIO and architect the firm’s intelligence layer — the system that determines *how we see the market*. This is not a quant role. This is not a reporting role. This is not a research-for-research’s-sake role. This is a **Human-in-the-Loop AI role**: building the *Iron Man suit* — an exoskeleton that allows a human trader to process **100× more information** without losing intuition, discretion, or speed. This role sits at the center of our trading advantage. ## **The Problem** Markets are noisy. Alpha comes from synthesis. We already have access to the data. The bottleneck is **human bandwidth**. Too much time is spent: - Refreshing feeds - Reading things that don’t matter - Manually stitching together signal across sources Latency kills insight. Insight kills latency. ## **The Mission** **Build the nervous system of the firm.**You will own the entire loop: **Ingestion → LLM synthesis → Trader decision** Your mandate is simple and brutal: **Reduce time-to-insight across the entire trading operation.** If you surface a signal 10 minutes faster than the street, we win. If you automate away four hours of reading, we double our capacity. ## **What You Will Own** **1. Information Pipelines (Kill the “F5” Key)** You will design and maintain automated pipelines that ingest, clean, and normalize: - News, filings, earnings calls, macro releases - Social and sentiment data - Alternative and proprietary datasets Your job is to replace manual refresh workflows with **push-based alerts** that surface *only what matters* — mapped directly to: - Watchlists - Live positions - Risk exposure Reliability matters. Latency matters. Silence matters. **2. AI-Augmented Research (A Live Analyst on the Desk)** You will deploy LLM-powered systems to summarize, extract, compare, and reason over: - 10-Ks, 10-Qs, earnings calls, central bank minutes - Sell-side research and internal notes You will build **“chat with our data”** tools that allow traders to query proprietary research in natural language. Example: “Summarize the bear case from the last three earnings calls.” You will also create tools for fast discretionary back-testing: “How did this asset behave during the last three macro shocks of this type?” **3. Signal & Noise Filtering (Protect Attention)** You will build the filters that decide *what breaks through*. - Sentiment and relevance scoring - Entity recognition that maps events to exposure - Dashboards that surface **regimes, anomalies, and dislocations** — not vanity metrics If something matters, it should scream. If it doesn’t, it should disappear. **4. Infrastructure, Security & Reliability (Own the Stack)** You own the compute layer that runs the intelligence system. - Cloud and/or local GPU infrastructure - Vector databases and retrieval systems - Data-privacy-first architectures (local models where required) You choose the architecture. You ship what runs fastest and breaks least. Our proprietary data and strategies do not leak. Ever. ## **Who You Are** **You Are a Founder-Hacker** You have likely: - Built your own projects - Traded your own account - Worked in a high-stakes startup, prop shop, or family office You hate waiting for permission. You like tools that get used *today*. **You Speak Python** - Python (Pandas, NumPy) is non-negotiable - You have hands-on experience with: - LLMs and RAG architectures - Prompting, evaluation, and failure modes - Vector databases (Pinecone, Milvus, FAISS, etc.) - Orchestration (LangChain, Airflow, or custom agents) - REST & WebSocket APIs (market data, news, internal tools) - Lightweight internal UIs (Streamlit, Dash, Retool, etc.) You know the difference between an LLM demo and a production system. You care about latency, hallucinations, and context windows. **You Get Markets (Enough)**You don’t need to be a portfolio manager — but you understand that: Financial data is messy, adversarial, and time-sensitive Narratives are not facts A 10-K is not a blog post You are comfortable making judgment calls under uncertainty. ## **The Deeter Mindset** **Co-Ownership Mentality** This is not a role for staff augmentation or support work. We are looking for a technical partner who thinks in outcomes, edge, and shared P&L — and wants to win in a competitive, adversarial market. We optimize for people who value: **Pragmatism over theory**You ship things that work by market open — not in six months. **Discretionary empathy**Human judgment is the final decision engine. Your systems amplify it; they do not replace it. **Founder-level ownership**You treat this system like your P&L depends on it — because it does. ## **Why This Role** 1. **Zero Latency** You sit with the decision-makers. You push code. We use it the same day. 2. **Real Impact** You are not optimizing ad clicks. You are building the intelligence engine for a $500M portfolio. 3. **Sovereignty** No middle management. No reporting theater. Just results. ## **To Apply** Do **not** send a cover letter. Send: - A GitHub - A project you’re proud of - Or a short note on how you would architect a real-time news filter for a trading desk We care about how you think and what you build.