Praneeth Tota
Sunnyvale, CA · Staff Engineer · AI Systems · Distributed Infrastructure

Building production-grade AI systems that learn from failure — without retraining

10+ years industry engineering, 6 years doctoral research — PhD in Algorithms & Game Theory.

Previously Tech Lead at Apple (global pricing infrastructure, contract via Infosys) and Walmart (last-mile delivery, 4,000+ stores, contract via Infosys).

Creator of AUA Framework v1.0 — a production framework for adaptive, utility-governed multi-model LLM systems. Routes prompts to specialist models, scores outputs, arbitrates conflicts, and prevents repeated errors across sessions without retraining.

What are you looking for?

Three audiences, three paths. Go directly to what's relevant.

🧑‍💼
Recruiters & Hiring Managers

10+ years software engineering on critical systems at Apple and Walmart. Hands-on delivery at scale, team leadership, cross-functional ownership.

⚙️
Staff+ Engineers & Interviewers

Distributed systems design, AI infrastructure, control plane architecture. Real systems at real scale with real failure modes.

🧠
Technical Peers & Curious Readers

Formal framework grounded in mechanism design and utility theory. VCG arbitration, Lyapunov-stable dynamics, validated simulation results.

Three things worth understanding

Research · 2026

AUA Framework v1.0

A utility-governed correction framework for deployed LLMs. Errors detected, stored as verified corrections, prevented from recurring — continuously, without retraining. Validated on physical hardware (RTX 4090): +43.3pp routing correctness gain over no-routing (p = 0.0003, d = 1.02). 69.6% repeated-error reduction across 5 calibration cycles. Ships as a production framework: REST API, CLI, Python SDK, Chat UI, Prometheus metrics, plugin system.

Full documentation site
Apple · Contract via Infosys · 2022–Present

Global Pricing Platform

Owned end-to-end design and evolution of the pricing platform powering Apple product pricing across all regions — high-stakes launch events, strict correctness and latency requirements. Led AWS migration reducing annual downtime from ~48 hours to ~4 hours. Validation frameworks cut production incidents by over 90%.

Full details in resume
Walmart · Contract via Infosys · 2019–2022

Last-Mile Delivery Platform

Technical lead on delivery infrastructure supporting 4,000+ Walmart stores under 2-hour SLA. Observability-driven workflows cut incident response from 45 to 20 minutes. Diagnosed Cassandra consistency failures, implemented root-cause fixes. Led team of 4 engineers maintaining 24/7 reliability.

Full details in resume

What I build

I design systems that enforce correctness and reliability in production environments — particularly in AI systems where failure modes are non-trivial and the cost of a wrong answer is field-dependent.

AI infrastructure & control layers
Adaptive feedback loops, LLM control planes, calibration pipelines, utility-governed deployment
Distributed backend systems
Microservices, event-driven architecture, Kafka/Flink pipelines, fault isolation at scale
Pricing & data platforms
High-correctness pricing infrastructure, OLAP pipelines, ETL validation, reconciliation at scale
Observability & reliability
Incident response systems, production debugging, SLA enforcement, Prometheus/Grafana/ELK

Background: PhD in Computer Science (Algorithms & Game Theory) — formally proved Price of Anarchy bounds in multi-agent resource allocation, the same mathematical structure as reward misalignment in RL. The research behind AUA Framework applies that foundation directly: VCG mechanism design for arbitration, Lyapunov stability analysis for behavioral dynamics, Kalman-optimal EMA for confidence tracking.