Miguel Ingram | AI Research Engineer
You might notice my contribution graph is unusually active. My workflow is defined by:
- 📊 High-Velocity Iteration: ~2365 commits/year
-
🔬 Granular Observability: Every change is tracked
-
⚡ Atomic Commits: Driving automated research pipelines
I treat development as a continuous experiment where each commit generates data.
I operate as an independent lab to bypass the "Golden Handcuffs" of corporate IP capture. My goal is not just research output, but Operational Sovereignty—owning the compute, the code, and the discoveries.
My background is rooted in high-stakes engineering. I started in the US Navy as a Nuclear Electronics Technician, where I developed an obsessive focus on:
- ⚙️ Systems Reliability: Zero-tolerance for failure in critical systems
- 🔍 Diagnostic Precision: Root cause analysis under pressure
- 📐 Operational Rigor: Translating Navy discipline into code
|
A pedagogical operating system using Runtime AST Mutation (surgical bug injection at the semantic level) and Shadow Worktrees (isolated Git environments for validation) to stress-test engineering skills. |
Research proving the "Compositional Gap" (models learn parts perfectly but fail to compose them) in Transformers. Validated against 302+ fine-tuned models. |
🧬 RNA Structure Prediction (Research Artifact)
A 99k LOC pipeline for 3D RNA folding implementing peer-reviewed methods:
- Stage B: TorsionBERT (Bernard et al., 2025)
- Stage C: MP-NeRF parallel kinematics
- Stage D: AlphaFold 3 diffusion refinement (Nature 2024) → Evolving toward SE(3) equivariant diffusion
Objective: Applying geometric constraints to biological foundation models outside of corporate walled gardens.
Stack: PyTorch Lightning • Hydra Config • 82% Test Coverage
I'm engineering a closed-loop "Cognitive and Operational Exoskeleton"—a system that quantifies human output across physical, technical, and cognitive domains to validate the Architect-Implementer-Verifier (AIV) model of human-AI symbiosis.
🎧 Listen to the System Audit (38m) An automated forensic analysis of the Cultivation Operating System, dissecting the AIV Model, Cognitive Debt mitigation, and the infrastructure behind the 14k+ commit history since 2019.
The goal: scale individual productivity by orders of magnitude while actively mitigating cognitive debt through systematic measurement and intervention. Status: [Internal Tool]
"Systematic Intervention on Biomechanical Limiters Under Environmental Stress"
Status: Data Collection Complete (103 Days). (Active Research Artifact).
|
|
868 activities • 4066 km • 767 hours • 480 active days
Lifetime running totals, automatically updated daily from Strava 🔥 API
Status: Allocating cycles for high-leverage Forensic Code Audits and specialized R&D contracts.
Objective: Funding next-generation compute acquisition (DGX Spark & H100 NVLink Cluster).




