Hi, I'm fooSynaptic

ML engineer focused on LLM post-training, alignment / RL loops, and inference tooling — the glue that turns research ideas into something you can run, measure, and ship.

I care about the assumptions under the code: simulation and tuning help, but they don't replace understanding the update rule.

---

Focus

• Post-training & optimization loops (GRPO / SAO-style asynchronous rollouts; one-factor ablation of RL mechanisms)
• Fast inference paths when serving stacks lag research realities (PEFT / LoRA merge; prefix KV reuse)
• Frontier model mechanisms (MLA, MoE, DSA, DSpark) as readable maps, not vibes
• Practical OSS hygiene — know when a fix is already upstream before opening a PR

Languages I reach for most: Python · C++ · Shell

---

Recently shipping

Repo	What it is
jev-any-llm	Wrap any instruct LLM into Jev-mode prediction — with KV-share branching so shared prefixes reuse one prefill
kimi3-align	Open one-factor ablation testbed for K3-style RL mechanisms on AReaL
deepseek-mechanism-atlas	Bidirectional wiki + mdBook of DeepSeek V1→V4 mechanisms (MLA / MoE / DSA / DSpark). Docs site

Selected tools

Repo	What it is
peft-adapter-merger	Offline PEFT adapter → dense merge, for reliable inference when serving frameworks trail latest LoRA / PEFT
Single-rollout-async-Optimization	Unofficial re-implementation of SAO (Single-rollout Asynchronous Optimization)
upstream-fix-gate	GO/STOP gate before opening OSS PRs — detect fixes already shipped upstream via gh
py_pcalg	Lightweight PC algorithm for causal skeleton discovery — pip + CLI + graph viz

Older course / trial repos stay in the account for history; they are not what I am building toward now.

---

GitHub Stats

---

Motto

fooSynaptic — foo as in foo / bar: learning means solving for the unknown; Synaptic for intelligence wired like synapses.

Contact

• Blog: foosynaptic.github.io
• GitHub: @fooSynaptic