Moinuddin Shaik
Before that, video work for 200+ clients. I was 15 when I started.
Different domains, different kinds of proof: competitive, commercial, published.
11
14
15
20
Now
Badminton
State-level competition. Student Nationals silver in doubles, 2018. Stopped during lockdown.
Code
Started in 10th class, 2020.
Video
Editing for clients on Fiverr, then direct. 200+ clients over roughly five years.
Amazon
Applied Scientist Intern. Production LLM systems over millions of feedback records.
Now
B.Tech CS (AI & ML), 9.09/10. Building AI systems, products, and research.
11
Badminton
State-level competition. Student Nationals silver in doubles, 2018. Stopped during lockdown.
14
Code
Started in 10th class, 2020.
15
Video
Editing for clients on Fiverr, then direct. 200+ clients over roughly five years.
20
Amazon
Applied Scientist Intern. Production LLM systems over millions of feedback records.
Now
Now
B.Tech CS (AI & ML), 9.09/10. Building AI systems, products, and research.
Four in depth: production LLM systems at Amazon, first-author taxonomy research, citation-first retrieval, and a durable multi-agent workflow. Each one starts with the operating constraint—not the model name.
A self-calibrating applied AI system that turns large-scale customer feedback into explainable three-level taxonomies—without a scientist hand-tuning every new domain.
0.74
F1 vs. 0.71 baseline
<18h
onboarding cycle
2.7×
faster extraction
An AI-native production studio that turns research papers and documentation into educational videos—built as a workspace to direct, not a prompt box to wait on.
8
departments, one orchestrator
3-step
generate / evaluate / revise loop
SSE
resumable execution stream
A citation-first document system that turns unstructured files into searchable, operational knowledge—without hiding retrieval behind a chat box.
10K+
documents designed for
<800ms
retrieval latency
<3s
end-to-end response
Research on measuring hierarchical structure and robust LLM classification—from hundreds to thousands of categories.
621–5K
category scale studied
25×
lower reported cost
2
2026 submissions
Current work focuses on hierarchy quality, classification at large label scales, and the cost of reliable decisions. Submission status is stated plainly.
First author
A framework for evaluating how automated systems construct hierarchical structure from raw feedback, including analysis of a subtle duplication failure mode in human-built taxonomies.
Third author
A scale-aware classification study spanning 621 to 5,000 categories, matching a frontier model's reported accuracy at 25× lower cost.
Cinematic video editing and motion graphics for clients worldwide. Fiverr first, then direct — across roughly five years, alongside school and then college.
I was 15, which is too young to hold a seller account — it needs legal documents I did not have — so I worked under my sister's. Fiverr first, then direct clients as the work grew. None of it was assigned to me: I found the work, taught myself the craft, and delivered to a brief on a deadline for people who were paying. Scope, revisions, and clients across timezones taught me the parts of building that have nothing to do with code, years before I had a job title. The gig is still up and I take the occasional project, but AI is the work now.
Six of the 117.
United States, Australia, United Kingdom, Mexico. Including a 4.3 — the average is 4.9, and it would be easy to show only fives.
Silver.
I picked up badminton at 11 and competed at state level before taking a doubles silver at Student Nationals in 2018. Lockdown ended it, but it was the first thing I got properly obsessed with — and the first time getting good at something was measured by someone other than me.
Jan–Jun 2026
Bengaluru
Amazon · RBS Sciences
Owned a self-calibrating knowledge-extraction system from problem framing to production, then expanded into autonomous taxonomy generation and grounded metric explainability. Cut onboarding from five to seven days to under 18 hours, reached 0.74 F1 against a 0.71 manual baseline, and accelerated extraction 2.7× at 53% lower cost.
May–Jul 2025
Hyderabad
Intel · Unnati
Optimized real-time video enhancement for low-resource devices: 20% clearer output, a 30% smaller model, and 35% faster CPU inference without a dedicated GPU.
These are not keyword buckets. They are the parts of AI systems where I like to make ambiguity explicit and performance measurable.
Grounded outputs, calibrated failure, provenance, and operator-visible evidence.
Layout-aware ingestion, ranking, hybrid search, citations, and measurable recall.
Task-specific metrics, structural error analysis, cost-quality tradeoffs, and judge reliability.
Typed pipelines, asynchronous execution, observability, reproducible experiments, and model serving.
Durable work boundaries, retries, idempotency, queues, consistency, and failure recovery.
Versioned APIs, data contracts, authentication, lifecycle design, and production safety.
Problem framing, baselines, efficient inference, on-device deployment, and product feedback loops.
Building evaluation into the workflow so quality changes the product, not just a dashboard.
Public repositories include product code, typed APIs, tests, CI, deployment notes, and explicit limitations—not only screenshots.
11
public repositories
Citation-first RAG, semantic search, and grounded QA for operational documents.
Agentic research assistant that verifies its own citations, with measured trade-offs.
On-device waste classification with React Native and TensorFlow Lite.
One page covering experience, education, selected systems, research, and technical foundations.
Moinuddin Shaik · Résumé · PDF
Open full screen ↗