# Full Profile: Christopher Mejo, Ph.D.
**AI Research Scientist — LLM & Reinforcement Learning Systems**
- Email: work@cmejo.com
- Location: New York, NY
- Website: https://cmejo.com
- GitHub: https://github.com/cmejo
- LinkedIn: https://linkedin.com/in/cmejo
- Security Clearance: Active Top Secret / Sensitive Compartmented Information (TS/SCI)

---

## Executive Summary
AI research scientist and engineer with 15+ years taking machine learning systems from research prototype to production, holding a Ph.D. in Physics from Harvard (quantum information and complexity theory, completed at age 21) and postdoctoral training at MIT under Dr. Scott Aaronson.
Core focus: Large Language Model (LLM) systems — retrieval-augmented generation, fine-tuning, agentic reasoning — and reinforcement learning for sequential decision-making.
Solo-architected AI Scholar, an open-source RAG platform indexing 15M+ research papers (Google Summer of Code Fellow, 2024 and 2025). Shipped production LLM and RL systems in finance and a federated computer-vision system in healthcare. Comfortable owning systems end-to-end in small teams.

---

## Professional Experience

### Senior Data Scientist — Braverock LLC (Nov 2021 – Present)
- Architected a reinforcement-learning trading system (deep Q-networks, actor-critic methods) for multi-asset portfolio optimization, improving Sharpe ratio 23% over prior strategy on microsecond-latency data, validated across both backtested and live trading.
- Built transformer-based market-prediction models for financial time series, deployed in a pipeline processing 1B+ quote-data points/day at microsecond-scale latency.
- Built a RAG system over 10M+ financial documents (SEC filings, earnings calls, research reports) for signal extraction, using custom embeddings and a production vector database.
- Directed distributed pretraining and instruction fine-tuning pipelines for domain-specific 7B–14B parameter foundation models (DeepSpeed ZeRO-3, custom CUDA memory-mapped tokenizers), reducing per-step wall-clock latency 38%.
- Implemented dynamic Heterogeneous Graph Attention Networks (HAN) modeling multi-asset cross-correlations across tick, intraday, and multi-week horizons, improving regime-shift detection accuracy 19%.
- Designed and deployed an enterprise crypto basis-yield strategy delivering an out-of-sample, beta-neutral Sharpe ratio of 2.63 on $60M+ in institutional AUM.
- Developed proprietary quantitative strategies achieving a 6.2 Sharpe ratio (Sortino 8.4, max drawdown <1.9%) over a continuous 24-month live deployment window, executing billions of dollars in trading volume across holding periods from sub-second to multi-month.
- Strategy scope spanned quantitative multi-exchange, multi-asset market making; treasury management; regime-aware risk-managed momentum; and multi- and single-asset statistical arbitrage / structural mean-reversion market-neutral strategies.

### Senior Data Scientist — Cushion AI (Apr 2021 – Oct 2021)
- Built CushionNLP, a proprietary BERT-based transformer system for bank-fee negotiation, driving a 30% increase in successful negotiations and $2M+ in additional annual customer savings.
- Architected a transaction-categorization system (ensemble methods, graph neural networks) achieving 95% accuracy across 80M+ transactions from ~300 financial institutions, cutting processing time 50% versus the prior rule-based (non-ML) system.

### Research Data Scientist — RapidRads AI (2019 – Feb 2021)
- Built a computer-vision system (EfficientNet-B7 backbone) for COVID-19 detection from chest X-rays, achieving 0.972 ROC-AUC with 96.5% sensitivity and 98.2% specificity (1.8% false-positive rate, 3.5% false-negative rate) under differential privacy (ε = 1.2, δ = 10^-5) across a federated-learning deployment spanning 4 medical institutions.
- Combined multiple CNN architectures via ensemble methods with Grad-CAM visualization and uncertainty quantification for clinical validation.

### Senior Data Scientist & Cybersecurity Researcher — Confiant (2017 – 2019)
- Built threat-detection models (Isolation Forest, One-Class SVM, LSTM autoencoders) and GraphSAGE-based botnet-detection models identifying coordinated attacks across 1B+ daily ad requests.

### CEO & Lead Architect — Cronode (2011 – 2017)
- Built a recommendation engine (TF-IDF, cosine similarity, collaborative filtering) and custom Solr ranking configuration.

---

## Concurrent Research & Fellowships

### Visiting Research Scientist (part-time) — Sandia National Laboratories & Joint Google Quantum AI Initiative (2020 – 2024)
- Senior researcher in a joint research initiative with Google Quantum AI developing variational quantum eigensolvers for lattice gauge theory calculations.
- Designed quantum error correction protocols (surface codes) and resource-estimation frameworks for near-term quantum devices (Qiskit, Cirq).
- Ran distributed quantum many-body simulations (50+ qubits) on SLURM-managed clusters; contributed to DARPA and IARPA research proposals.
- Active Top Secret/Sensitive Compartmented Information (TS/SCI) security clearance.

### Google Summer of Code Fellow (2024 & 2025) — AI Scholar Project — R Foundation for Statistical Computing (2024–2025)
- Architected AI Scholar (github.com/cmejo/AI_Scholar), combining dense (FAISS), sparse (BM25), and graph-based (Neo4j) retrieval for scientific literature search across 15M+ research papers.
- Built an adaptive retrieval system using meta-learning to personalize search strategy, improving relevance 40% (NDCG@10) over baseline on a 200-query held-out set with citation-derived relevance judgments.
- Implemented chain-of-thought reasoning over the RAG pipeline with explicit source attribution and calibrated uncertainty (Monte Carlo dropout, evidential deep learning).
- Built a dynamic knowledge-graph construction pipeline using named entity recognition (spaCy, BERT-NER) and relation extraction to connect concepts, authors, and methodologies.

---

## Education
- **Postdoctoral Fellowship, Theoretical Computer Science** — Massachusetts Institute of Technology (MIT) (2010–2012). Faculty Advisor: Dr. Scott Aaronson.
- **Ph.D. in Physics** — Harvard University (2007–2010). Focus: Quantum information and complexity theory. Degree completed at age 21.
- **B.S. in Physics and Computer Science** — Harvard University (2004–2007).

---

## Technical Skills
- **Languages**: Python, C, C++, R, Assembly, Java, JavaScript
- **ML / Deep Learning**: PyTorch, TensorFlow, JAX, Transformers, Reinforcement Learning (DQN, actor-critic, MAML)
- **LLM / RAG**: Retrieval-augmented generation (dense, sparse, and graph-based retrieval), FAISS, BM25, Neo4j knowledge graphs, custom embedding models, vector databases, LangChain, LangGraph, vLLM, named entity recognition (spaCy, BERT-NER), chain-of-thought and agentic reasoning, uncertainty quantification (Monte Carlo dropout, evidential deep learning), retrieval evaluation (NDCG, MRR)
- **MLOps**: MLflow, Kubeflow, Apache Airflow, Docker, Kubernetes, Ray, Weights & Biases
- **Quantum Computing**: Qiskit, Cirq
- **Distributed / HPC**: SLURM, multi-GPU distributed training

---

## Target Roles & Value Offered
- Roles: Principal / Staff AI Scientist, Lead LLM Systems Researcher, Quant ML Researcher, Head of AI Research.
- Key Strengths: Bridging cutting-edge foundational theory (Harvard PhD / MIT CS postdoc) with battle-tested low-latency production engineering (1B+ events/day, microsecond execution, distributed GPU training, multi-modal enterprise systems).
- Willing to relocate / remote / hybrid in New York, NY or major tech hubs.