Berwyn
Software Engineer — I build distributed systems, ML pipelines, and AI tooling.
#about
Hi, I'm Berwyn — a software engineer and CS student at Stanford
(BS/MS, expected 2028), working across distributed systems, machine
learning, and AI infrastructure.
I like building things from scratch — a Raft-inspired in-memory cache in Go,
ML pipelines processing 10B+ records, and AI visibility tooling that cut report
generation from hours to minutes.
#projects
memflux
A distributed in-memory cache built from scratch: TTL eviction, gRPC replication, and Raft-inspired leader election with sub-500ms failover. 56M reads/sec single-threaded, zero allocations per read.
Buzzly
AI-powered marketing content generator. Python backend on the OpenAI Agents SDK with custom orchestration and safety guardrails, plus a fault-tolerant async multimedia pipeline (Heygen, DALL·E, Runway, Whisper).
#education
Stanford University — BS & MS, Computer Science (GPA 3.80)
Expected Jun 2028Coursework: ML for Algorithmic Trading, NLP with Deep Learning, Deep Learning for Computer Vision, Reinforcement Learning, Design & Analysis of Algorithms.
#experience
Software Engineer @ The Prompting Company
Jun 2026 — PresentBuilt a Go CLI (15+ subcommands) for AI bot-traffic analytics, and an end-to-end AI visibility pipeline across ChatGPT, Perplexity, and Gemini — Temporal fan-out + Tinybird aggregation cut report generation from 3 hours to under 5 minutes.
Machine Learning Engineer @ Telkomsel
Jun 2025 — Aug 2025Designed a hybrid model-serving architecture (in-house inference + AWS) that replaced Vertex AI and cut annual infra cost by 90%. Processed 10B+ clickstream records in BigQuery and built CatBoost ranking pipelines end-to-end.
ML Research Assistant @ Stanford Materials Science & Engineering
Jun 2024 — Aug 2024Built a PyTorch + C++ pipeline showing lightweight Moment Tensor Potentials reach 90% of SOTA equivariant-NN accuracy. Co-authored a peer-reviewed paper on a split-model framework with a 10x gain in energy/force prediction accuracy.