Cyma Systems
AI-native engineer matching with Graph RAG, deterministic eligibility filters, and graph + semantic retrieval.

hey, i’m vidith. i’ve been building a lot around AI and backend systems lately and i’m starting to go deeper into multi-agent orchestration. most of the stuff i’ve worked on is up on my github ↗ if you want to look around.
AI-native engineer matching with Graph RAG, deterministic eligibility filters, and graph + semantic retrieval.
Multi-tenant construction SaaS used by two companies, with Gemini multimodal document ingestion and role-based workflows.
TypeScript · PostgreSQL · Advisory locks · Vitest · Model-based testing · npm package
Next.js · TypeScript · FastAPI · PostgreSQL · pgvector · Redis · Docker
Next.js · TypeScript · OAuth · Gmail API · Microsoft Graph · MCP
Where well-understood AI work should actually run, and why model size turned out to be the wrong variable.
IoT telemetry and alerting with MQTT/PostgreSQL, plus an LLM compliance copilot for restaurant operations.
FastAPI + OpenAI backend services supporting 11,000+ users across production courses.
GCP/BigQuery data engineering, Python automation, and GA4 migration work.
React + Spring Boot product work with SQL automation that cut commission processing from days to about an hour.
Three upstream fixes: per-task conversation isolation for concurrent AgentRearrange runs, MajorityVoting configuration validation, and agent removal by name in SocialAlgorithms.
3 merged PRs · shipped in v15 “Akira” ↗ · named release contributor
Added deterministic state assertions and fail-closed, read-only HTTP probes for verifying external system state, with structured passed / failed / error evidence and credential-safe configuration.
merged upstream · documented in the SDK README ↗
Scheduling, pthread concurrency, memory allocation, and low-level C.
Lexing, parsing, ASTs, scoping, and static type checking.