AI Engineering Lead + Head of Frontend Platform

I build production AI systems where the model is one component inside a larger product surface: retrieval infrastructure, agent runtimes, guardrails, evals, traces, cost controls, permissions, and operator-facing interfaces.

I have 20 years of software delivery experience across product, frontend, backend, and platform work. The current focus is making AI useful in real financial operations, where answers need evidence, workflows need state, interfaces need to be operable, and failure modes need to be visible.

I currently hold a dual role: I lead AI engineering across three production systems, and I lead the frontend platform behind an enterprise product surface. The names are internal; the engineering shape is the signal.

BIFROST

Document intelligence and retrieval infrastructure for real-world inputs: ingestion quality gates, semantic chunking, semantic and visual retrieval, pgvector/HNSW search, caching, source-quality summaries, analytics, and honest no-answer behavior.

Python · FastAPI · PostgreSQL · pgvector · Docling · PyTorch · Transformers

ORVIAN

AI workflow runtime for B2B operational workflows: protected multi-tenant APIs, context assembly, durable memory, deterministic/cached/full-LLM execution tiers, run events, idempotency, queue processing, scheduled maintenance, and human-review metadata.

TypeScript · Hono · PostgreSQL · Drizzle · Supabase · queues · scheduled jobs

Polaris

Internal AI assistant product integrating BIFROST retrieval with MONARCH guardrails, cached safety-to-retrieval handoff, citations, streaming UX, visual query support, query analytics, and suggestion revalidation.

Next.js · Vercel AI SDK · BIFROST · MONARCH · Drizzle ORM · PostgreSQL

• Retrieval infrastructure: ingestion, chunking, metadata, semantic search, visual search, citations, source quality, and no-answer paths.
• Agent runtimes: state, context assembly, tool use, execution tiers, idempotency, queues, handoffs, and human review.
• AI guardrails: prompt-injection checks, context sanitization, confidence thresholds, escalation rules, and audit trails.
• Product reliability: latency, token/cost metadata, observability, eval loops, regression checks, and operator-facing UX.
• Full-stack delivery: product judgment across interfaces, APIs, databases, deployment, and operational constraints.

gommage

Policy-as-code permission harness for AI coding agents: deterministic rules instead of prompt-only safety.

traceframe

Local-first trace recorder and inspector for AI agent workflows: append-only run evidence, Codex/OMX-style hook ingestion, ledger indexing, HTML reports, and CI-verified smoke tests.

vestig

Runtime-agnostic structured logging with context propagation, observability primitives, and automatic PII sanitization.

The public line is becoming more coherent: permission boundaries, inspectable traces, structured logs, eval gates, retrieval quality, tool contracts, and human-review boundaries for production AI systems.

Most AI systems fail for ordinary engineering reasons: weak context boundaries, vague orchestration, no eval path, poor observability, and no clear ownership when the model is wrong.

I prefer systems where the behavior can be inspected. Retrieval should show its evidence. Agents should expose state and stopping conditions. Guardrails should be explicit. Cost and latency should be visible. Human review should be part of the design, not a late-stage patch.

The hard part is not getting a prototype to answer once. The hard part is making the system reliable enough that people can use it when the input is messy, the context is incomplete, and the model is unsure.

petruarakiss.com · LinkedIn · GitHub · contact@petruarakiss.com