Ryan Rochmanofenna

Forward Deployed AI Engineer · Jakarta, Indonesia

Four-stage LangGraph pipeline structuring 500k+ Indonesian legal and corporate PDFs into a knowledge graph, plus the cross-store RAG service and Go telemetry behind the in-house DekaLLM platform.

See the ingestion pipeline and a retrieval trace →

Software Engineer · NYC

Generative Manim pipeline pairing GPT-4o and Claude to produce narrated CFA educational video, with async streaming that cut perceived latency 70% and failover holding 99.9% uptime.

See the pipeline run, end to end →

Early Software Engineer · Remote

Containerized ManimGL on Modal and Fly.io GPUs with an async batch pipeline: 5x throughput, 35% lower cost, 1080p renders in 38s at p99.

See Manim rendering at this scale →

Software Engineer, Trading Execution Systems · Remote

Systematic trading system running end to end across three strategies on a 2GB droplet: real-time ingestion, phase-gated model promotion, out-of-sample evaluation infrastructure, and a Rust hot path for decision-time inference.

Software Engineer Intern, Platform · NYC

Real-time order-tracking backend on AWS Lambda serving 10k+ concurrent sessions at 99.9% uptime.

MCP server that gates AI-coding-agent actions against a per-codebase learned risk model, then seals each action into a tamper-evident, SHA-256 hash-chained receipt with a verifying CLI.

Verify a real receipt in your browser →

Multi-tenant restaurant platform: one Next.js app serving many restaurants by subdomain, with AI-generated storefronts and live QR-to-cashier ordering into the ESB point-of-sale.

Read the writeup →

iOS ordering and loyalty app for my parents' F&B group in Jakarta, spanning several restaurant brands. Also the first live tenant on Sajian.

See the app →