About Me
I’m an AI engineer specializing in agentic LLM systems, human-centered AI, and optimization—with a strong background in UX and front-end web development. I have 10+ years of experience shipping production web applications.
Notable Achievements
- Reduced ADHD diagnosis time by 80% by building an AI-assisted mental healthcare platform from ideation through early clinical testing.
- Achieved 1500x+ speedups (1 day → 1 minute) by implementing custom GPU parallelization using Python, NumPy, Numba, and CUDA/C.
- Came up with six novel top-performing solutions within my first 100 questions on LeetCode (write-up), including a top 10% runtime and memory solution for a LeetCode Hard (example: Trapping Rain Water — beats 91% runtime and 93% memory).
- 2024 AI Conference Hack Day HackrChallenge Winner (LinkedIn announcement).
- 10+ years shipping production web applications; delivered 30+ end-to-end digital projects across product strategy, UX, and engineering.
Experiences
- Leading development of end-to-end agentic AI workflows supporting the full lifecycle of sponsored research.
- Focused on searching, matching, and filtering RFPs (requests for proposals), saving thousands of hours for top academics.
- Skills: AI Agents, Google Cloud Platform (GCP), Google Vertex, A2UI.
- Led development of an AI-assisted mental healthcare platform from ideation through early clinical testing.
- Architected low-latency conversational workflows integrating databases, cloud LLMs, and text-to-speech.
- Designed safety guardrails (input validation, PHI filtering, topic boundaries, human fallback routing).
- Designed key components for a scientific theory mining platform for interpretable few-shot learning.
- Implemented GPU-parallel K-means (Python/NumPy/Numba/CUDA C) achieving 1500x+ speedups.
- Built RAG pipelines and LLM-powered interfaces using LlamaIndex and vector databases.
- Drove product strategy, architecture, and commercialization for a private equity research platform.
- Delivered 30+ end-to-end digital projects (UX, branding, A/B testing, iterative optimization).
- Grew user base 20x by launching user feedback initiatives and data-driven onboarding experiments.
- Built a collaborative community platform from scratch with retention-focused UX improvements.
- Led end-to-end front-end development of a cross-institutional fMRI research platform.
- Implemented an Angular architecture enabling shared access to a multi-institution scan repository.
- Visiting Fellow studying AI Interpretability.
Projects
A small selection of personal and product work.
Exobrain - Personal cognitive scaffolding system for working with AI assistants.
polishedscreenshot.com - Guidelines for documentation screenshots, written for humans and agents. Source for polishedscreenshot.com.
batched-kmeans-python-numba - Batched K-means on GPU with Numba and CUDA C. 1500x speedup over the CPU baseline. Companion to the Towards Data Science article.
Claude Code Skills
Claude Code skills I've built. More at github.com/Minda/skills.
- lowkey-viz — Markdown, TUI-style visualizations (distributions, timelines, correlations, outliers) that surface patterns and edges in any dataset.
- lumenscope — Maps a text onto 10 polar-opposite poles, scores 7 aspects per pole using exact quotes from the text, and surfaces meta-patterns.
- venn — ASCII Venn diagram comparing two or three sources: what only one claims, what they share, and where they contradict.
- creating-ascii-drawings — ASCII diagrams that make complex conceptual relationships tangible and memorable.
- de-ai — Rewrites AI-generated text to sound more human by removing telltale patterns and adding authentic detail.