/resume
Operator Record
Resume material is rendered as structured technical history, with a downloadable PDF for the canonical document.
summary
Junior at UC Irvine specializing in systems, ML inference, and performance engineering. Shipped production-grade work across AI accelerator kernel optimization, edge-cloud inference pipelines, and full-stack ML systems, with comfort spanning bare-metal OS internals through cloud infrastructure.
education
University of California, Irvine
B.S. Computer Science
Redmond, WA
Expected March - June 2027
GPA: 3.707
Coursework: Operating Systems, Computer Architecture, Data Structures & Algorithms, System Design, Machine Learning & AI, Embedded Systems, Networking, Parallel and Distributed Computing
skills
Languages
C | C++ | C# | Java | Python | TypeScript | Verilog | Assembly
AI / Inference
PyTorch | TorchScript | YOLO | Gemini API | latency analysis | memory layout | data movement
Systems / Tools
Linux | Git | Docker | QEMU | log analysis | trace analysis
Full-Stack
FastAPI | Next.js | PostgreSQL | REST APIs | Supabase
Cloud / Infra
AWS EC2 | S3 | SQS | RDS | Cognito | Amplify | DynamoDB | IAM | GitHub Actions CI
download
open pdfvaibhav.attre@gmail.com
https://github.com/VaibhavAttrehttps://www.linkedin.com/in/vaibhav-attre-89005024a/activities & awards
WebJam Hackathon (3rd Place): real-time AI PhotoBooth web app with computer vision and live filters.
AI Club: course planning assistant.
Video Game Design Club: enemy AI behaviors and modular ability system.
AWS Club @ UCI: event operations lead.
experience
GoFlyy
ML Systems Intern (Computer Vision Pipeline)
January 2026 - Present
Irvine, CA
Designed and shipped an end-to-end clothing scan ML pipeline using Gemini for initial classification, with migration in progress to YOLO and a custom-model roadmap.
Validated behavior across 9 garment categories through 100+ targeted regression tests.
Architected the backend around FastAPI, Next.js, PostgreSQL on AWS RDS, S3 artifact storage, and an async job queue with traceable DB/API flows for auditability and debugging.
Integrated AWS Cognito, Amplify, and RDS while keeping ML scoring decoupled from ingestion through explicit schemas and API contracts.
UCI Undergrad Research
Researcher (ML Systems)
January 2026 - Present
Irvine, CA
Developing runtime guardrails for early-exit and split inference to maintain accuracy and latency QoS under variable wireless conditions.
Instrumented bailout-event logging across 1000+ runs to inform adaptive thresholding strategy.
TuriyamAI
Software Development Intern
June 2025 - September 2025
Redmond, WA
Improved throughput by 12% on selected Tenstorrent Blackhole kernels by optimizing NoC-aware data movement and specialized functional units.
Profiled 8 custom PyTorch operators, reduced kernel latency by 9% through tiling and double buffering, and cut stall time by 20% on targeted traces.
Built repeatable perf and validation harnesses with standardized configs, 60-run batches, regression checks, and summary output adopted across the team.
Idori
Software Development Intern
January 2025 - June 2025
Irvine, CA
Increased average FPS by 35% in constrained WebGL via GPU instancing, batching, and shader simplification, while reducing frame-time variance by 18%.
Eliminated race and crash incidents from 5+ per week to under 1 per week through a thread-safe producer-consumer pipeline built with mutexes and semaphores.
selected projects
TinyOS
RISC-V OS + disk-backed CoW filesystem | C, RISC-V, QEMU
2025 - Present
Built an RV64 OS from scratch with Sv39 page tables, page-fault handling, trap/syscall entry and return, timer-interrupt preemption, and round-robin scheduling.
Implemented a disk-backed copy-on-write filesystem with a VirtIO block driver, buffer cache, checksummed B-tree metadata, atomic commits, extent allocation, refcounted CoW writes, snapshots, subvolumes, and reflink clone.
Hardened reliability with invariant checks and 100+ iteration allocation and CoW stress loops to surface subtle memory bugs.
SentinelQ
Edge-cloud home surveillance | Python, C++, Next.js, FastAPI, PostgreSQL
March 2026
Built a low-cost surveillance platform on Qualcomm Arduino UNO Q handling 4+ simultaneous camera streams.
Designed an adaptive inference pipeline with a branchy ResNet that exits locally or escalates to cloud based on connectivity, camera quality, and latency constraints.
Shipped the full stack: Next.js frontend, FastAPI backend, PostgreSQL via Supabase, on-device HTTP debug endpoint, and an LLM assistant for text-command threat summaries.
Telemetry Platform
Cloud QEMU experiment runner | Python, QEMU, Docker, AWS
2026 - Present
Automated QEMU workload runs to collect kernel.log, run_meta.json, and metrics.json artifacts while diffing runs to catch regressions across 200+ runs and 10 workloads.
Built an AWS-backed queued runner: SQS jobs to EC2 workers running containerized QEMU, storing artifacts in S3 and indexed metadata in DynamoDB.
Added GitHub Actions CI for smoke and regression workloads on every push.