Yuvraj Shiv — EECS @ UC Berkeley

Overview

I'm an EECS undergrad at UC Berkeley who builds machine learning systems that run in production, not just in notebooks. At Altametrics I cut daily sales-forecast error from 12% to 5% across four national restaurant chains — McDonald's, Taco Bell, Jack in the Box, and Chipotle — covering thousands of locations.

Outside of ML, I like building systems from the ground up: a 10,000+ LOC relational database engine with a cost-based query optimizer and ARIES-style crash recovery, a Git-like distributed version control system, and a 32-bit RISC-V CPU designed from the gate level.

I'm looking for Summer 2027 internships in machine learning, ML infrastructure, and backend/systems engineering. Happy to talk to anyone working on forecasting, ML systems, or database internals — yuvrajshiv@berkeley.edu.

Experience

Machine Learning Intern — Altametrics

March 2025 – Present · Costa Mesa, CA

Architected and deployed the production ML forecasting pipeline predicting daily sales for 4 national QSR chains across thousands of restaurant locations.
Cut forecast error from 12% to 5% (~58% relative improvement), driving tighter labor scheduling, lower food cost, and reduced overtime spend.
Engineered 30+ predictive features from weather, zip-code demographics, holidays, seasonality, school calendars, and national events, plus macroeconomic indicators (CPI, unemployment).
Own the end-to-end model lifecycle: preprocessing, feature engineering, cross-validated hyperparameter tuning, and back-testing over 2+ years of history.

Python · XGBoost · LightGBM · CatBoost · Random Forest · Prophet · TimesFM

Software Engineering Intern — MAG Services

May 2023 – August 2023 · Costa Mesa, CA

Built Java-based data-analysis tooling processing 12+ weeks of restaurant sales and labor-cost data to surface scheduling inefficiencies across 20+ locations.
Designed a labor-optimization playbook adopted by operations managers that cut overtime pay and eliminated unprofitable shifts.
Delivered 15+ analytical reports and stakeholder presentations across 20+ recurring meetings.

Java · SQL · Excel

Semiconductor Research Intern — COSMOS, UC Santa Cruz

July 2023 – August 2023 · Santa Cruz, CA

Selected for the competitive COSMOS summer research program at the Baskin School of Engineering; researched device physics of LEDs, transistors, solar cells, and memristors.
Led a 4-person team to design and prototype a 4-watt solar panel.
Authored and presented a technical project on memristor applications in neuromorphic computing and ML hardware acceleration.

Projects

RookieDB — Relational Database Management System

CS 186 · Fall 2025

Implemented core RDBMS internals in a 10,000+ LOC SQL engine: B+ tree indices, hash joins, sort-merge joins, and grace hash joins.
Built a cost-based query optimizer using dynamic programming over join orderings and selectivity estimates.
Engineered ARIES-style crash recovery (write-ahead logging, ACID rollback/redo) and multi-granularity two-phase locking.

Java · SQL

Pac-Man AI Agents

CS 188 Artificial Intelligence · Spring 2026

Implemented graph search (DFS, BFS, UCS, A* with custom heuristics) and multi-agent adversarial search (minimax, alpha-beta pruning, expectimax).
Built reinforcement-learning agents using value iteration, Q-learning, and approximate Q-learning with feature-based linear function approximation.
Trained Bayesian networks and Hidden Markov Models for probabilistic ghost localization under noisy sensor observations.
Implemented a Naive Bayes digit classifier from scratch, reaching 80%+ accuracy on held-out data.

Python · NumPy

Gitlet — Distributed Version Control System

CS 61B · Spring 2025

Built a Git-like version control system from scratch (~3,000 LOC) supporting init, add, commit, log, branch, checkout, merge, reset, and remote operations.
Used SHA-1 content-addressable storage and implemented 3-way merge with conflict detection via lowest-common-ancestor traversal.

Java

CS61CPU — 32-bit RISC-V Processor

CS 61C · June – August 2025

Designed a 32-bit RISC-V CPU in Logisim with an ALU (add, sub, shifts, bitwise, multiply), 32-register file, and immediate generator.
Built a single-cycle datapath supporting 30+ R/I/S/B/U/J-type instructions via ROM-based control logic.
Automated correctness verification with Bash-driven testbenches against custom RISC-V assembly programs.

Logisim · RISC-V Assembly · Bash

MeshEdit — Geometry Processing

CS 184 Computer Graphics · Write-up ↗

Implemented de Casteljau subdivision for Bézier curves and separable evaluation for Bézier surfaces.
Built area-weighted vertex normals, edge flip, and edge split on a half-edge mesh, then composed them into loop subdivision.

C++ · Half-edge data structure

Treble Boost Guitar Effects Pedal

EECS 16B · Spring 2026

Hand-derived the symbolic transfer function for an op-amp bandpass topology; selected RC values targeting ~10× voltage gain across a 300 Hz – 15 kHz passband.
Prototyped with bypass capacitors and soft-clipping diodes for noise rejection; validated against SPICE simulations and oscilloscope / network analyzer measurements driving a live guitar amp.

Analog circuit design · SPICE

Education

UC Berkeley — B.S. Electrical Engineering & Computer Science

Expected May 2028

Relevant coursework: Artificial Intelligence & Machine Learning (CS 188), Database Systems (CS 186), Computer Networks (CS 168), Data Structures & Algorithms (CS 61B), Computer Architecture (CS 61C), Structure & Interpretation of Programs (CS 61A), Designing Information Devices & Systems I / II (EECS 16A / 16B).

Riverside City College / Norco College — Dual & Concurrent Enrollment

August 2021 – May 2024

Completed 40+ units of advanced math, computer science, and physics coursework concurrently with high school; maintained a 4.0 GPA.