I build production-grade intelligent systems: from fine-tuning LLMs and orchestrating multi-step AI agents to shipping decentralized ML at scale on blockchain infrastructure.
At Digital Currency Group, I architected VCScout-AI, a LangChain-based agentic RAG pipeline powered by Qwen-7B and Weaviate that analyzed 300+ startup deals and directly drove $450K+ in AI-informed investments. Previously at Intel, I led TensorFlow CPU optimization, earning the Division Achievement Award and shipping TF 2.11 to 10,000+ downloads in a single day.
My research on federated LoRA fine-tuning (adaptive phase-switching, 40.5% measured communication savings) is under review at TMLR. I write on HackerNoon, speak at EthCC and NEARCON, and guest-lecture at universities on production LLMs.
| $450K+ Impact | 7x Top Performer | EthCC + NEARCON | 2 Papers Under Review |
|---|---|---|---|
| AI-informed investments via VCScout-AI | FLock AI Arena | Invited Speaker | TMLR + PeerJ CS (Q1) |
LLMs and Fine-Tuning
Agentic AI and RAG
ML Infrastructure and MLOps
Cloud and Languages
Federated and Decentralized ML
| Project | Description | Stack |
|---|---|---|
| VCScout-AI | Agentic RAG pipeline analyzing 300+ startup deals, driving $450K+ in AI-informed investments at DCG | LangChain · Qwen-7B · Weaviate · GCP |
| Decentralized AI on FLock | Federated learning models ranked Top Performer 7x across distinct AI Arena tasks | Federated Learning · PyTorch · Blockchain |
| Federated LoRA Fine-Tuning | Adaptive phase-switching achieving 40.5% communication savings — under review at TMLR | PyTorch · TinyLlama · LLaMA-3.2 |
| finance-news-chat-app | Conversational search interface for financial news with keyword retrieval and query intent classification | Flask · Python · RAG |
| Brain Tumor Detection | 12-layer TensorFlow CNN achieving 0.96 sensitivity on MRI classification | TensorFlow · CNN · Medical AI |
| Amazon Recommender System | Collaborative filtering with PCA, Agglomerative Clustering, and Matrix Factorization (MAE 0.096) | Python · Scikit-learn · PCA |
- Adaptive Phase-Switching for Communication-Efficient Federated LoRA Fine-Tuning — Under review at Transactions on Machine Learning Research (TMLR) · OpenReview #9042
- Fine-Tuning Large Language Models in Resource-Constrained Environments: Methods and Trade-offs — Under review at PeerJ Computer Science (Q1)
- QLoRA Explained: The Memory Compression Breakthrough — HackerNoon, May 2026
| Event | Topic | Year |
|---|---|---|
| NEARCON 2026, San Francisco | Lightning talk: Training 7B LLMs on edge devices | 2026 |
| EthCC, Brussels | Invited talk: Decentralized AI and federated learning on FLock | 2024 |
| Guest Lecture, VIT-AP University | Production AI: LLMs, LoRA/QLoRA, and real-world deployment | 2026 |
| Intel AIA Division Achievement Award | TensorFlow on Windows — 10K downloads on day one | 2022 |
| FLock AI Arena | Top Performer 7x across distinct tasks, 1st place in two major hackathons | 2024 |
Senior AI/ML Engineer · Open to impactful roles · jerryadamsfranklin.com · Fairport, NY