Work
Doozie Software Solutions
Working on LLM and full-stack problems. Most of my time goes into web applications, product features, and practical AI workflows.
Now
What I am focused on right now.
Research & Systems
MendSpeech (personal lab)
Building MendSpeech, a real-time dictation system, in my own personal lab. The question I'm chasing is narrow: can the system make a transcript easier to read without changing what the speaker said, and what would that cost in latency? Shipped so far: a seeded acoustic-damage suite, a frozen 30-utterance benchmark, and Wav2Vec2 recognition with greedy CTC decoding and alignment written from first principles. In progress: cache-aware streaming ASR, confidence calibrated against correctness, and a conservative editor that is allowed to format a transcript but never to rewrite it.
Personal
Quiet baseline
Calm music, reading habits, health, and how I try to think about attention and distraction in a noisy world.
Writing
Build notes and essays
Writing here when something needs a longer explanation. Shorter notes may also go to Hashnode.
Public updates
Short notes on X
I use X for quick updates on AI tools, automation ideas, coding, and small lessons from building.