Jatin Jassi
By theme

Technical, AI & ML

Applied GenAI shipped end to end, PyTorch deep learning on high-performance computing, and a comparative cloud data-platform evaluation, with reliability, cost and evaluation treated as first-class concerns.

Technical · AI / ML Featured

Vergence: Automated Data Quality and Reconciliation Pipeline

A controls-first pipeline that measures where two independent sources disagree, with one language-model step held behind a human approval gate.

7 controls registered, 7 run
9 exceptions traced to 5 planted defects19 columns resolved from the registry on the second run, zero model callsByte-identical exception register across three mapping paths16 tests, all passing
Pydantic pytest SQL / Cloud Data Azure GenAI / LLM
View project
Technical · AI / ML Featured

Agentic AI Workflow for Scalable Document Generation

A nine-agent LangGraph orchestration that generates tailored application documents with a hard truthfulness constraint.

50% lower estimated per-row cost
Batch time cut from 37 hours to under 8 hours for 200 applications9 agents across 12 modules
Azure GenAI / LLM
View project
Technical · AI / ML

Retrieval-Augmented Azure OpenAI Assistant

An internal knowledge assistant that answers staff questions on products, policies and procedures in natural language.

Replaced manual portal searches with natural-language access
Reduced follow-up questions from frontline staff
Azure GenAI / LLM
View project
Technical · AI / ML

AI-Driven Melanoma Detection

A PyTorch deep-learning model for classifying dermoscopic melanoma images, built for early-detection use in healthcare.

+12% accuracy on imbalanced data
93% precision25% faster convergence15,000+ images, 88% GPU utilisation
Python / ML / DL
View project
Technical · AI / ML

Generative Adversarial Network for Image Synthesis

A Wasserstein GAN that generates high-resolution synthetic faces to address data scarcity.

15% FID stability gain vs baseline
40% faster training50,000+ images90% model efficiency in constrained settings
Python / ML / DL
View project
Technical · AI / ML

Sparse Random Projection for Medical Image Analysis

A feature-extraction pipeline that compresses high-dimensional medical imaging while keeping diagnostic signal.

65% dimensionality reduction
35% faster than PCA10,000+ images standardisedScales to 3D with under 10% code change
Python / ML / DL
View project
Technical · AI / ML Featured

Cloud Data Platform Evaluation

Benchmarked Snowflake, Azure Synapse and Amazon Redshift for AI and ML workloads and large-scale analytics.

8-dimension scoring matrix
4 years of sales data98% faster integration on Snowflake vs Redshift40% lower query latency on Synapse serverless3x faster model training via Snowpark
SQL / Cloud Data
View project

Contact

Full-time unrestricted working rights, Subclass 485 Temporary Graduate Visa, valid until January 2028.