Applied AI Engineer

Simran Malik

An applied AI engineer building AI-powered solutions across AI agents, search & retrieval, model training and fine-tuning. As a NeurIPS 2026 co-author, skilled in evaluating AI systems and the sources they cite. Previously, a software engineer for 4+ years at Google shipping critical high-volume distributed systems.

I combine my powers of backend software engineering at scale, AI-powered solutions & evaluations and research-level rigorous experimentation to build AI systems which will work on a bad day.

Resume

4+ yrsGoogle

NeurIPS2026 co-author

2Patents in Preparation

M.S.CS · AI, UC San Diego

Work

Oct 2026 - Present

Existential Robotics Lab · UCSD

Applied AI Researcher

Oct 2026 - Present

Enabling robots to answer natural language questions about the world around them.

Jun - Sept 2026 2+ months

Brain Corp · San Diego

Intern · Applied AI · Software Engineering

2+ months · Jun - Sept 2026

Built a knowledge graph and an agent that answer natural-language questions about a robot-built 3D store map

Sep 2025 – Nov 2026 1+ year

WukLab · UCSD

Applied AI Researcher

1+ year · Sep 2025 – Nov 2026

Built the evaluation pipeline behind SourceBench (NeurIPS 2026), measuring the quality of sources cited by 12 AI systems.

Jul 2021 – Aug 2025 4+ years

Google · India

Software Engineer

4+ years · Jul 2021 – Aug 2025

Shipping user-facing features, making payment systems more reliable, resolving stuck user funds, collaborating cross-functionally across Business, Tax, Legal, Product teams and third-party vendors.

Jan – Jul 2021 6 months

Apple · India

Intern · Software Engineering

6 months · Jan – Jul 2021

Designed and built a data management portal for the Apple Ads platform, saving 10+ engineering hours per week.

Publications

NeurIPS 2026 · Evaluations & Datasets TrackAccepted

SourceBench: Can AI Answers Reference Quality Web Sources?

H. Jin*, S. Liu*, Y. Li*, S. Malik, Y. Zhang (*equal contribution)

Evaluated web sources cited by 12 AI systems on 8 quality metrics across 100 diverse questions. Proposed the research direction and the human-labeled ground truth, and built the evaluation pipeline behind the main results.

Led the Google vs. AI-search study: top AI systems cite sources largely absent from Google's top results (16% overlap for GPT-5 vs. 55% for Tavily). Also caught a prompt-bias bug that would have skewed every system's results.

Paper

Projects

RAG-based Q&A

Multi-stage retrieval for technical documentation under a 2,000-token budget; ranked 3rd of 141 teams with 0.978 correctness.

LangChain · FAISS · HuggingFace

Supervised Fine-Tuning for Natural Language Schema Linking

Four-stage pipeline of fine-tuned adapters on Qwen2.5-Coder, raising schema-linking score from 0.097 to 0.627.

PyTorch · PEFT · Qwen2.5

Small Language Model Training

Trained a 98M-parameter Transformer on 10B tokens; ranked 5th, and quadrupled GPU utilization to 47%.

PyTorch · nanoGPT · RTX 5090

Growth rate predictor

Predicting a population's growth rate from genetic data with an MLP and a CNN trained on simulated populations.

Python · PyTorch · msprime

Law-monade

Hackathon build that turns a live personal-injury case into a one-screen brief, with every fact linked to its source.

Python · FastAPI · Claude / Gemini

Pupper Potter: An interactive trivia with the Mini Pupper robot

A Harry Potter–themed game on a quadruped robot, coordinated by a distributed ROS2 state machine.

ROS2 · Python · Piper TTS

Blog All posts ↗

To autonomize, or not to autonomize?

Build the discretion to choose between a deterministic system, an LLM application and an agentic system.

NOV 2026 · 6 min read

Data, data and then some more data

Data can be the difference between an AI solution that seems to work and one that does. Data can be the reason your ML model doesn't work…

NOV 2026 · 2 min read

Life Away from keyboard

Usually, you can find me detecting edge cases where my AI system would fail, figuring out why, and iterating over solutions. Otherwise, you will find me curled up with a cosy murder mystery and a hot cup of tea!

Murder Mysteries

I have read all Agatha Christie books :)

Chai

Sipped in peace with the morning sun

Knitting & Crocheting

Crafted 4 mufflers, 1 pair of gloves, 1 crochet top while listening to audiobooks

Contact

Have a question? Interested in collaborating? Get in touch.

I usually reply within 24 hours. You can also find me here:

Resume