Mohammed Shaik
I build AI systems that are accurate, fast, and cheap to run in production.
CS student at UT Dallas (graduating Dec 2026), most recently building LLM-powered tools at Staples and ML feature pipelines at Intuit. I care about the parts of AI engineering that don't show up in a demo — eval harnesses that catch regressions, retrieval that's actually grounded, and inference that stays fast and cheap at scale.
Resume
Work Experience
AI Engineer Intern
Staples
- Built and deployed an LLM service in Python and FastAPI that handles internal operations requests end to end: it reads the request, calls internal APIs for the data it needs, and completes the action, replacing manual triage that cost 15 hours a week.
- Designed the tool-calling layer that lets the model choose among 8 internal APIs, validating arguments against a schema, retrying invalid calls, and falling back to a deterministic path, holding success above 99% across ~8,000 requests a day.
- Added retrieval over internal policy documents with hybrid keyword and vector search plus reranking, raising accuracy from 78% to 94% on a 500-question labeled set and cutting hallucinated values by 60%.
- Built an evaluation harness with a labeled golden set and model-graded scoring that runs in CI on every prompt or model change, blocking merges that regress accuracy, latency, or cost.
- Cut inference cost by 35% by routing simple requests to a smaller model and caching repeated lookups, and added per-request token and latency tracing to find the expensive paths; deployed with Docker on Kubernetes.
Software Engineer Intern, Data Platform
Intuit
- Built Python components of a streaming pipeline handling about 40 million events a day from Kafka, turning raw events into user-session features for downstream machine learning models.
- Wrote features into a shared store so training and serving read the same definitions across 10+ event types, and built a monitoring tool that flags when the pipeline falls behind and recommends how much to scale up, cutting peak-hour delays by 30%.
- Moved 15+ jobs to AWS with S3 and Redshift, scheduled them with Airflow, and wrote validation scripts confirming no records were lost, improving query speed on training data by 25%.
Projects
01 / 07
Skills
Languages
Machine Learning
LLMs & Generative AI
Evaluation & Optimization
Vector & Databases
MLOps & Cloud
Data Engineering
Certifications
PCEP – Certified Entry-Level Python Programmer
Python Institute
Microsoft Certified: Azure Fundamentals
Microsoft
Microsoft Certified: Azure Data Fundamentals
Microsoft
Recent Events
HackUTD
HackathonNov 2025 · Richardson, TX
Location Discovery Platform
At UT Dallas's flagship 24-hour hackathon, built a location-discovery platform with a team of four that scrapes TikTok and Instagram for trending places, uses AI to interpret natural-language search and generate location descriptions, and runs geospatial search over PostgreSQL/PostGIS.
01 / 02