Applied Scientist · AI Systems Engineer
Building reliable AI systems.
I design retrieval, extraction, evaluation, and ML infrastructure that turn messy data into decisions people can inspect and trust.
Hyderabad, India
Open to applied science + AI systems roles
0.74 F1
Automated taxonomy generation
10K+ docs
Citation-first retrieval
<300ms
Offline mobile inference
01 · Selected work
Four case studies spanning production LLM workflows, taxonomy research, retrieval infrastructure, and on-device ML. Each one starts with the operating constraint—not the model name.
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
12. Term and termination
Renews annually unless either party gives 60 days’ written notice.
01 · Agreement.pdf
Page 18 · 96%
02 · Renewal schedule
Page 4 · 89%
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
Unstructured feedback
Induced hierarchy
F1 · 0.74Research 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
A mobile waste-classification prototype that brings fast, offline guidance to resource-constrained devices.
82%
model accuracy
<300ms
on-device inference
−60%
model size
Recyclable
82% confidence
<300ms
−60%
Offline
02 · Publications
Current work focuses on hierarchy quality, classification at large label scales, and the cost of reliable decisions. Submission status is stated plainly.
First author
Abstract
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
Abstract
A scale-aware classification study spanning 621 to 5,000 categories, matching a frontier model's reported accuracy at 25× lower cost.
03 · Experience
01 · Feb–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.
02 · 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.
04 · Technical interests
These are not keyword buckets. They are the parts of AI systems where I like to make ambiguity explicit and performance measurable.
01
Grounded outputs, calibrated failure, provenance, and operator-visible evidence.
02
Layout-aware ingestion, ranking, hybrid search, citations, and measurable recall.
03
Task-specific metrics, structural error analysis, cost-quality tradeoffs, and judge reliability.
04
Typed pipelines, asynchronous execution, observability, reproducible experiments, and model serving.
05
Durable work boundaries, retries, idempotency, queues, consistency, and failure recovery.
06
Versioned APIs, data contracts, authentication, lifecycle design, and production safety.
07
Problem framing, baselines, efficient inference, on-device deployment, and product feedback loops.
NOW
Building evaluation into the workflow so quality changes the product, not just a dashboard.
05 · Field notes
A future home for technical notes drawn from systems I have actually built and evaluated. No trend summaries; only concrete decisions and evidence.
Why provenance has to survive ingestion, retrieval, generation, and the final interaction—not live in a hidden log.
Draft · In progress
A field note on duplicated concepts, boundary ambiguity, and evaluation that respects structure.
Draft · In progress
Designing a read-only product tour that shows complete workflows without accepting data or invoking a model.
Draft · In progress
06 · Open source
Public repositories include product code, typed APIs, tests, CI, deployment notes, and explicit limitations—not only screenshots.
GitHub snapshot · Jul 2026
13
public repositories
Citation-first RAG, semantic search, and grounded QA for operational documents.
On-device waste classification with React Native and TensorFlow Lite.
A small Python utility for reducing image datasets while preserving training utility.
07 · Résumé
One page covering experience, education, selected systems, research, and technical foundations.
Moinuddin Shaik · Résumé · PDF
Open full screen ↗