4 Years of Engineering Experience in 4 Months
How AI shifted the job from line-by-line coding toward orchestration, review, and delivery management.
Writing on AI systems, OCR, LLMs, and product delivery
This archive organizes the site by topic instead of chronology so the strongest internal links are explicit and the older essays are easier to interpret in context.
Some posts from 2019-2021 are preserved as historical snapshots. They are still useful, but they reflect the state of the field at the time they were written.
AI Product Delivery
Field reports on shipping with coding agents, choosing stacks, and keeping quality high while moving faster.
From Vibe Coding to Compound Engineering
A design-philosophy essay on why AI-assisted software still needs structure, memory, and explicit process.
Lessons From Vibe Coding an iOS App
What actually mattered while building a real-time voice translation app with AI tools across app and backend layers.
LLM Systems + Research Notes
Experiments and essays on context length, token efficiency, prompting, and practical LLM engineering.
Token Compression
An experiment on compressing multiple tokens into one representation to reduce attention waste.
Long Pythia
Extending transformer context length and documenting the tradeoffs that show up in practice.
Getting the Most Out of Pre-trained Models
An older but still useful note on prompting, question answering, and adapting pre-trained NLP models.
OCR + Computer Vision
Mobile UI parsing, OCR evaluation, and machine learning features that shipped into real testing products.
Autify OCR vs. Google OCR
A detailed OCR benchmark focused on rendered mobile UI text and the failure cases that matter.
Machine Learning Features in Autify for Mobile
How visual regression, self-healing, and app exploration fit together in a mobile testing workflow.
Solving Automated App Navigation
Behavior cloning, heatmaps, and uncertainty-aware tap prediction for mobile automation.
Applying Modern Deep Learning in Autify
Production-oriented experiments using segmentation, graph learning, and RL inside a testing company.
Cloud + ML Infrastructure
Practical notes on deploying models, choosing infrastructure, and understanding cloud tradeoffs.
Deploy ML Models on Google Cloud Run
A hands-on deployment tutorial with a clear note that the hosted demo is no longer live.
Cloud Run
A high-level introduction to why serverless containers mattered for ML application delivery.
Deep Learning in Cloud
A cost-oriented comparison of the cloud options that were available for deep learning at the time.
Historical Snapshots
Older essays that are still worth keeping online, but should be read as time-bound snapshots rather than current guidance.