Profile: honest snapshot (update quarterly)¶
A short, honest self-assessment against the north star, reviewed every quarter in the monthly review. It exists so that decisions at the roadmap gates are made from evidence, not mood.
As of: September 2026 · North star: core AI (frontier-lab research engineer) and/or AI founder. Decide by evidence at the month-6 and month-9 gates (ROADMAP, founder decision gate).
Assets¶
- About 2 years of shipped production ML: Transformer OCR for government utility automation at Sujanix (accuracy gains, INT8 ONNX deployment on AWS Lambda); privacy-first RAG (FinSentinelAI); an agent with memory (ECHOME).
- Founder and operator experience: SpaceDrift, a sole proprietorship run for 16 months (Aug 2024 – Dec 2025), with scoping, pricing, delivery and paid contractors. Rare, and useful for both tracks.
- Breadth across CV, NLP, LLMs and deployment; Kaggle Expert; an RTX 5060 for real experiments.
Info
Every number attached to these assets (accuracy, cost reduction, uptime, time saved) must come from your own records before it appears on a résumé or in an interview. The open items are listed in resume/guide.md and stories.md.
Gaps (the curriculum exists to close these)¶
- Depth below the API: no from-scratch implementations of training, RL, inference kernels or scaling experiments yet. → labs 01–17.
- No research artifacts: no reproductions, ablations with error bars, or publications. → module 08, portfolio.
- Coding-interview fluency: DSA self-rated intermediate. → 3–4 h/week maintenance.
- Signal and network: a lesser-known college and employer mean public work and open source must supply the brand. → public-presence.md.
- Relocation constraints for US roles. → visa-and-relocation.md.
Positioning (rewrite each quarter as evidence lands)¶
One sentence for each track, used in outreach, the résumé summary line and "tell me about yourself". Only claim what you can link.
Research engineer: "ML engineer who has shipped document AI to production and is rebuilding the
LLM stack from scratch: <latest lab or experiment, with link>."
Founder: "Engineer-operator who has sold and delivered ML services and shipped production OCR;
now validating <problem> with <n> customer interviews and <n> pilots."
Leading indicators (track these; no invented probabilities)¶
Labs passing · experiment write-ups published · open-source PRs merged · mock-interview pass rate · recruiter replies per 10 targeted applications · paid pilots or LOIs (founder track). If these rise, outcomes follow; if they stall, change the plan, not the metrics.
| indicator | Sep 2026 | Dec 2026 | Mar 2027 | Jun 2027 | Sep 2027 |
|---|---|---|---|---|---|
| labs passing (of 17) | |||||
| write-ups published | |||||
| OSS PRs merged | |||||
| mock pass rate (last 5) | |||||
| replies per 10 targeted applications | |||||
| paid pilots / LOIs |