Chiara Infra, Data, and Security
Technical substance and business pragmatism.
Who We Are
Through Chiara, Founder & Principal Mariyam Mirza provides technical guidance to growth-stage startups and enterprises facing high-stakes infrastructure scaling, data science/ML, and security and privacy decisions.
Mariyam's academic and professional expertise is evaluating and improving the architectural efficacy of planetary-scale systems — speed, throughput, cybersecurity, and cost — using real, messy, incomplete data.
Having worked at three sector-forming companies — Akamai Technologies (Content Distribution Networks), BitSight Technologies (Cybersecurity Ratings), and Coalition Inc. (Active Cyberinsurance) — Mariyam revels in solving new, ill-defined, “no one else does this” problems. She is currently building a privacy-preserving cell-phone-based system that detects when humans become incapacitated and gets help to their pets.
She has a B.S.E., magna cum laude, from Princeton University, and a Ph.D. in Computer Science from the University of Wisconsin-Madison. Her Ph.D. work was one of the earliest to apply machine learning to problems in computer networks.
What we do
1. Infrastructure Scale & Cost Management
What We Do: Design compute and storage architectures that balance technical needs and business goals.
Why It Matters: Infrastructure matches unit economics. Prevents runaway cloud spend.
Engagement Scenarios
- Your current data and compute infra outgrew your user base. We architect the correct infra and stack for the next phase of your growth.
- Your infra is — contrary to projections — struggling to keep up with load, or your cloud bill is higher than expected. We diagnose why (load balance issues? underutilized compute or memory? suboptimal data partitioning?), and recommend fixes.
- You want to enhance your product, but the infra and dev work to build new pipelines will cost too much. Are there ways to capture most of the product enhancement for significantly less cost, or do it in phases to minimize risk? We answer these questions.
Completed Projects
- Implemented PySpark pipeline for a 100 billion rows / day datafeed at BitSight. Optimized various stages – some memory-bound, others CPU-bound – using low-level, undocumented Spark system parameters to achieve cost-performance sweet spot.
- Akamai's flagship ecommerce product’s content download speed was much slower than expected, putting significant revenue at risk. Analyzed data from multiple sources across the Akamai cloud to identify the cause. Led to fixes and desired performance and revenue.
- A critical internal Akamai network of 500+ servers, responsible for maintaining ground truth records, was running out of capacity faster than projected, resulting in unexpected costs. Added telemetry, analyzed resource utilization of complex, custom workload spanning the entire network, and resolved the issue by identifying significant load imbalance.
2. Security Data Science and Systemic Risk
What We Do: Move organizations beyond patchworks of solutions and tackle security holistically, addressing it as the distributed systems and data-at-scale problem that it is.
Why It Matters: As AI use dramatically expands attack surfaces, designing scalable, cost-effective telemetry and remediation strategies is essential to understanding security posture and building defense strategies.
Engagement Scenarios
- Evaluating internal and external data sources for use in understanding overall security posture.
- Architecting scalable data pipelines for high-volume telemetry, balancing visibility against compute and storage costs.
- Combining disparate data sources using modeling techniques to understand your organization's complete security posture and develop prioritized improvement plans.
Completed Projects
- Used internal and third party datasets to (a) correlate malicious activities (scanning, bots, …) with ransomware and business email compromise (BEC), (b) compare ransomware and BEC risk on major cloud platforms. Results informed actuarial risk selection and cyber insurance pricing at Coalition.
- Analyzed an Internet-wide 100 billion rows / day dataset, combined with BitSight internal and third-party datasets as needed to understand coverage of traffic, and resultant security signals, across the entire IPv4 address space.
3. Defending Revenue via Competitive Benchmarking
What We Do: Empirical benchmarking of key product attributes such as latency, throughput, efficient resource utilization, and completeness of coverage.
Why It Matters: Helps win and retain customers using quantitative evidence to counter marketing noise. Prevents blind spots and competitive surprises.
Sample analysis available upon request.
Engagement Scenarios
- We develop reproducible competitive benchmarking methodologies for internal use to drive product strategy and maintain competitiveness.
- We publish peer-review-quality technical reports that establish your product’s performance or coverage advantage.
Completed Projects
- Demonstrated that BitSight’s cybersecurity telemetry predicted ransomware incidents, making BitSight the first ratings provider to establish market legitimacy in this manner.
- Crafted Akamai’s technical rebuttal by conducting rigorous measurements in response to a competitor’s highly public whitepaper claiming better CDN performance.
4. End-to-End Data System Architecture
What We Do: Set data-intensive ML projects up for production success.
Why It Matters: The majority of ML/AI projects never make it to production, because models are the focus, and the whole data ecosystem — scalable and affordable data pipelines, changing data distributions, production deployment upkeep — is bolted on at the end, often unsuccessfully. We flip this script, developing holistic, balanced architectures focused on continued production viability.
Engagement Scenarios
- Evaluating whether proposed model architectures are viable under real-world infrastructure constraints, data availability, and latency budgets before millions are spent on training and deployment.
- Architecting robust data ingestion, feature verification, and evaluation harnesses so models perform reliably without silent data drift or unexpected failure modes.
Completed Projects
- Customers need to understand vendor-of-vendor (4th party) relationships to evaluate supply chains’ cyber resilience. Identified pragmatic and cost-effective data sources / pipelines, metrics, and models for BitSight’s industry-leading 4th party product.
- BitSight was expanding its offering beyond security ratings to exposure, critical for navigating major security incidents. Evaluated technical design space, and defined 6-9 months of cross-functional (data science, security research, engineering) technical work for this initiative.
Get in touch
Please direct inquiries to mariyam@chiaratech.dev
© 2026 Chiara Infra, Data & Security. All rights reserved. Design adapted from HTML5 UP.