Irkutsk (UTC+8) · Junior+ · Open to remote and hybrid
Async task queues · REST APIs · AI agents and RAG
Focus: reliability, observability, tests.
Problem: reliable background job processing without losing tasks on worker crashes Implemented:
|
Problem: ingesting a 340k-record snapshot (110 MB) under a 256 MB memory limit Implemented:
|
Implemented:
|
Problem: answering questions from documents without sending data outside Pipeline: PDF → chunks → embeddings → ChromaDB → top-3 → Qwen2.5
|
What it does: the LLM decides which tool to call Tools: calculator · weather · RAG search
|
What it does: thermal image classification — NORMAL / OVERHEAT / FAULT Accuracy: ~85–90% on validation
|
🌿 hololinked — Collaborator
Python framework for data acquisition and IoT
-
PR #191 (merged): attribute docstrings below property definitions — AST parser in
ThingMeta. Explicitdoc="..."wins; graceful fallback for REPL/notebooks. 4 tests. -
PR #200 (merged): MQTT WebSocket transport for publisher and consumer. TLS-enabled WebSocket listener,
wss:///ws://TD forms. 2 e2e tests. -
PR #205 (merged): MQTT v5 content type from properties in consumer. Safe
getattrfallback for paho/aiomqtt. Bumped version to 0.4.2.
Merged into main, included in releases v0.4.1 and v0.4.2.
Python Backend / AI Developer (Junior+ / Middle)
Remote · Hybrid · Relocation

