Ananya Salian

ML Researcher & AI Engineer · BSc Statistics, University of Melbourne

↓ Download CV

Skills

Programming & ML

PythonPyTorchTensorFlowScikit-learnHugging FacePyTorch-GeometricRDKit

Technical Areas

Diffusion ModelsPrivacy-preserving MLRAGEvals & Model CalibrationGNNsSemi-supervised LearningFine-tuningUncertainty Quantification

Research Interests

InterpretabilityModel robustnessStatistical evaluationComputational chemistryAI safety

Experience

AI Engineering Intern

Jan 2026 – Present

Affinda Group , Pathfindr Build Team & Internal Tools · Melbourne, Australia

  • Delivered Azure OpenAI + RAG workflow automation for a member-based enterprise serving ~1.7M customers/year; integrated with Microsoft SharePoint knowledge stores and CRM.
  • Architected an AI-assisted intake and fraud-screening workflow for an ops platform handling ~1K inbound cases/month, integrating into preexisting architecture.

Cybersecurity Intern

Dec 2023 – Present

Crysp Consulting · Melbourne, Australia

  • Contributed to a cybersecurity RFP (~$250K) for government clients in APAC & EMEA regions.
  • Conducted threat analysis and drafted reports for organisations in the medical and health sectors.

Research

FaceSafe: Privacy-Compliant Inpainting Pipeline for Scalable Image Datasets

First co-author

ICML 2025, DIG-BUG & WiML

  • First co-author; built a context-preserving inpainting pipeline enabling large-scale privacy-preserving facial dataset augmentation
  • Presented at ICML 2025 workshops: Data in Generative Models (DIG-BUG) & WiML
  • Mentored by a Meta FAIR researcher and an ML researcher at Brown University (Algoverse Collaboration)
Paper ↗

MoE Router Interpretability

Researcher
  • Studying failure modes in MoE routers as toy models of latent misalignment, focusing on hidden misaligned experts and deceptive routing modes

RAVEL: Receipt-Anchored Verifiable Evaluation Ledger

Sole author
  • Sole author; protocol for auditable, pre-label certify-or-abstain evals under covariate shift using PSIS-κ/ESS diagnostics and self-normalised importance sampling
  • Evaluated on 19 text, graph, and molecular benchmarks: ~3–13× reduction in test universes at fixed PPV floors
  • Converts many NO-GUARANTEE cohorts to certified PPV floors; runs on a single CPU machine

Selected Projects

RAG Evaluation Framework

rag, evaluation

Production eval pipeline for RAG workflows , Affinda Group

  • Delivered Azure OpenAI + RAG workflow automation integrated with SharePoint knowledge stores and CRM for a member-based enterprise serving ~1.7M customers/year
  • Architected an AI-assisted intake and fraud-screening workflow for an ops platform handling ~1K inbound cases/month
  • Prototyped additional AI workflows for internal operations across the Affinda tooling suite

Achievements & Programs

  • BlueDot Impact Technical AI Safety Course Certification
  • ARENA AI Safety Curriculum , through APAC Technical Alignment Research Accelerator (TARA)
  • Jane Street INSIGHT Trading Program Participant
  • Kliqtek Scholarship , HEX Youth Entrepreneurship Program: built a credit-scoring ML model (MAPE 1.88%) and pitched an ML-fintech startup; awarded full scholarship
  • Citadel Securities Trading Invitational , Top 20
  • UBS × Morgan Stanley Finance Accelerator , 3rd and 4th place

Leadership

Sponsorship Director · HackMelbourne2024–25
Publications Officer · Finance Students' Association2024
External Relations Officer · StartUp Link2024
Division Director · Interact Rotary International2020–24

Education

Bachelor of Science: Statistics & Stochastic Processes

Jan 2024 – Present

The University of Melbourne · Melbourne, Australia

Coursework: Algorithms & Data Structures, Linear Algebra, Corporate Finance, Machine Learning, Linear Statistical Models, Engineering Mechanics, Investments