Fraud detection
99.2%
recall
Booking Q3 engagements
// Dr. Aria Khan
Intelligence, shipped.
AI/ML engineer building intelligence that ships to production — LLM fine-tuning, computer vision, recommender systems and RAG at scale.
99.2%
best model recall
40ms
vision p99 latency
3M
docs in RAG index
8
models in production
Capabilities
What I build
→LLM Fine-tuning
→Computer Vision
→Recommender Systems
→RAG Pipelines
LLM Fine-tuning
Domain-adapted LLMs — SFT, LoRA/QLoRA, eval harnesses and guardrails, shipped behind a stable endpoint.
Computer Vision
Detection, segmentation and OCR pipelines tuned for real-time, low-latency inference on edge or GPU.
Recommender Systems
Two-tower retrieval and learned ranking that measurably lift engagement and revenue.
RAG Pipelines
Retrieval-augmented generation over millions of docs — chunking, embeddings, reranking and MLOps.
Engagements
Ways to work together
Scoped engagements — from a focused sprint to embedded delivery.
AI Audit
A focused two-week review of your data, models and infra — with a prioritized, build-ready roadmap.
$4000
Accepting now
Start this engagement
Most popular
Model Sprint
Six weeks to a shipped model — scoping, training, evaluation and a production endpoint your team owns.
$18000
Accepting now
Start this engagement
Embedded Delivery
Monthly retainer — I embed with your team to ship, monitor and maintain models end to end.
$40000
Waitlist only
Join waitlist
Featured work
Models in production
Selected systems — the metric that mattered, shipped and measured.
Real-time vision QA
40ms
latency
Enterprise search RAG
3M
docs
Talks
On the record
NeurIPS workshop talk
Signal
What teams say
Aria shipped a fraud model that actually held up in production — 99.2% recall with latency we could live with. Rare to find research depth and shipping discipline in one person.
Dr. Elena Vasquez
VP Engineering, FinShield
Our QA line went from manual to real-time vision in six weeks. 40ms inference, rock-solid. She left us with a clean endpoint and docs the team could own.
Marcus Reed
CTO, VisionWorks
The RAG system she built indexes 3M documents and retrieval finally feels relevant. Thoughtful evals, not vibes — that changed how we trust the output.
Priya Nair
Head of Search, Lumen
Embedded with us for two quarters and lifted our recommender CTR meaningfully. Genuinely the best applied-ML hire we have made.
James Okoro
Founder, DataForge
Currently available
Booking Q3 engagements
Open to applied-research consulting, model delivery and advisory work.
Discovery
Strategy call
30 minutes. Your problem, the data you have, and whether a model is the right call.
Contact
Start a conversation
Publications & links
Papers, talks & profiles
Availability
Studio hours
Monday
9:00 AM - 6:00 PM
Tuesday
9:00 AM - 6:00 PM
Wednesday
9:00 AM - 6:00 PM
Thursday
9:00 AM - 6:00 PM
Friday
9:00 AM - 1:00 PM
Saturday
Closed
Sunday
Closed
Lab
Where I am
Find me online