~/whoami → Surya Avala
Principal AI Systems Architect · 10 years across 4 regulated industries
"A decade of making ML systems survive production in healthcare and energy — where the plumbing matters more than the model."
Healthcare · Energy · Finance · Tech
Six measured outcomes across Healthcare, Energy & Recommender systems.
reduction in feature lead times across the GCP ML platform
Montu — Healthcare ML Platform (2023–2025)
Replaced sprawling notebook-to-production pipeline with Kubeflow + GitOps, turning sprint-length deploys into same-day ships.
DORA metrics, measured pre/post Kubeflow + GitOps platform migration
clinical NLP accuracy — outperforming Google Healthcare NLP by 14 points
Montu — Clinical Document Intelligence
Built custom clinical NLP pipeline outperforming Google Healthcare NLP on domain-specific medical text extraction.
Held-out clinician-labelled validation set, 5-fold cross-validation
average infrastructure cost reduction across 10k+ energy sites
Amber — Energy Forecasting Platform (2021–2023)
Designed event-driven FinOps architecture for 10k+ energy sites, eliminating redundant compute and cold-path waste.
GCP billing diff over 6-month rolling window post-FinOps refactor
clinician adoption of the prescription recommender (100k+ patients)
Montu — Two-Tower Recommender System
Hybrid two-tower recommender personalising prescriptions for 100k+ recurring patients with clinician-in-the-loop feedback.
CI [68.1% – 74.5%] · 95% CI bootstrapped over weekly cohorts
Active-user telemetry / total clinicians eligible, 90-day rolling window
clinician case reviews automated by the care quality assessment pipeline
Montu — Care Quality Assessment
Automated clinician case review pipeline combining structured extraction with quality scoring, freeing clinical hours.
Automated-review count / total reviews, monthly rolling
F1 PII redaction score on clinical log sanitisation
Montu — Privacy-by-Design Logging
Privacy-by-Design log sanitisation with tuned NER model, ensuring clinical inputs never leak PII to downstream systems.
Held-out clinician-annotated PII corpus, micro-averaged F1
Targeting Principal / Staff IC roles in regulated ML infrastructure.
Volunteering for Social Good · long-form writing on notes.
last verified
computed at build · not a marketing claim
Production agentic search engine — GraphRAG, Celery DLQ, Circuit Breakers, Kubernetes. Featured on the homepage as the flagship agentic system.
O(1) stdlib optimisation — zero GC thrashing, pure CPython heap/deque. Demonstrates first-principles algorithmic thinking against real workloads.
5 merged PRs across MLOps + data infra
Open linkRFC authorship · ML Guild mentoring · cross-functional Build vs Buy negotiation.
"Hired, onboarded and mentored teams of Data Scientists & ML Engineers across multiple orgs."
full leadership detail per role →
Open linkUsing ML to drive equitable outcomes — climate action, public-policy data, and technology for underserved communities.
"AI is the new electricity."
AI for Social Good · Climate · Rational Thinking