Currently leading AI engineering at Centric Software
Livin Mathew.
Solutions Architect and AI engineer. I build production GenAI platforms, multi-agent RAG, and serverless systems on AWS for enterprise clients.
- experience
- 11+ yrs
- scaled
- 10M+ users
- cost cut
- −35%
- audits
- first-pass
soc 2 · hipaa · hitrust
/now
Recently joined Galent as Lead AI Engineer, working with Centric Software on PLM AI projects. Focus areas: multi-agent systems and LLM tooling over PLM data (styles, BOMs, suppliers, materials, compliance).
Cochin, India · remote
Selected work
Things I've built & shipped
-
mideastpulse · live
MidEast Pulse
500+ DAUMiddle East news aggregator and pulse tracker. Curated regional coverage with real-time updates.
News Aggregationlive -
sage · 6 agents · 7 locales
GenAI knowledge platform at Sage
2024 — 2026Multi-agent RAG with semantic caching across 7+ locales, ingesting 500k+ documents nightly into a hybrid OpenSearch vector index. A 15-metric LLM eval framework cut inference cost by 35% and pulled customer query response time from 2 hrs to 10 min.
AWS Bedrock LangGraph OpenSearch Step Functionscase study -
soc 2 · hipaa · hitrust · first-pass
Clinical AI/ML at SiteRx
2020 — 2023End-to-end AI/ML pipeline classifying patient EHR documentation for clinical trial study matching. Architected HIPAA-compliant ingestion; passed SOC 2 and HITRUST audits on first attempt. Eliminated 95% legacy tech debt, cut infra cost by 35%.
AI/ML HIPAA HITRUST Serverlessprivate -
diagram → terraform
KloudArch
healthcare pilotsPlatform that converts architecture diagrams directly into deployable Terraform. Currently powering pilot clients in healthcare.
Terraform Go Canvas Healthcareprivate -
claude · codex · gemini
PitStop
open sourcemacOS menu bar app that tracks usage limits across Claude Code, Claude Desktop, OpenAI Codex and Google Gemini accounts, then switches between them in one click when one hits its rate limit. Reads each provider's credentials locally and swaps them in place.
Swift SwiftUI macOS Dev toolsgithub -
mic + system audio · on-device
MeetNote
fully localPrivacy-first meeting notes for macOS. Captures mic and system audio from a Teams/Zoom/Meet call, transcribes both on-device with Apple's SpeechAnalyzer, and summarizes with a local LM Studio model. Nothing touches the network — menu bar app and CLI over one engine.
Swift SpeechAnalyzer LM Studio Privacygithub
Experience
Eleven years of shipping.
-
2026 — now
Lead AI Engineer · Galent → Centric Software (remote)
- — Recently joined. Will be working on PLM AI projects: multi-agent systems and LLM tooling over Centric's PLM data (styles, BOMs, suppliers, materials, compliance).
-
2024 — 2026
Mobile Product Lead · TEKsystems → Sage (remote, UK)
- — Designed multi-agent RAG (FastAPI + LangGraph, 6 concurrent agents) with semantic caching, Claude integration, and multilingual support across 7+ locales.
- — Architected Bedrock-powered semantic search ingesting 500k+ documents nightly into an OpenSearch hybrid vector index.
- — Built a 15-metric LLM evaluation framework (groundedness, relevance, latency, token cost) that cut inference cost by 35%.
- — Reduced customer query response time from 2 hrs to ~10 min via automated RAG workflows.
- — Engineered distributed ingestion using Step Functions (10 concurrent maps), multi-region CDK stacks, and GitHub Actions CI/CD with OIDC federation.
- — Integrated OpenTelemetry observability. Achieved zero security findings during audit.
-
2023
Solutions Architect · ChiSquare Labs → SiteRx (remote, US)
- — Led the monolith to serverless migration (Lambda, DynamoDB, EventBridge); cut infrastructure cost by 35%.
- — Extended AI/ML data pipelines for patient EHR classification: added feature modules, improved model accuracy, scaled ingestion for growing trial volumes.
- — Built reusable CDK templates (logging, monitoring, security) adopted across 20+ engineers.
- — Directed SOC 2 compliance and secure-SDLC implementation across cross-functional teams.
-
2020 — 2023
Solutions Architect · Metric Tree Labs → SiteRx (remote, US)
- — Owned the architecture for end-to-end AI/ML data pipelines that classified EHRs for clinical trial study matching, from requirement scoping through production.
- — Architected HIPAA-compliant ingestion; passed HITRUST audit on first attempt.
- — Designed DR strategy with defined RTO/RPO targets and validated failover drills across multi-module enterprise systems.
- — Eliminated 95% legacy tech debt; lifted client revenue 40%; lifted developer productivity by 30%.
-
2019 — 2020
Software Architect · Riafy Technologies
- — Scaled Android applications to 10M+ downloads through backend modernization.
- — Improved legacy AngularJS platform performance by 90%; led design reviews and guided implementation teams.
-
2014 — 2019
CEO & Co-founder · Lmntrx Tech
- — Bootstrapped startup to acqui-hire exit; owned end-to-end architecture, hiring, delivery, and P&L.
- — Reduced operational cost by 25%; tripled traffic via scalable cloud architecture and growth optimization.
Stack
What I reach for.
Languages & frameworks
- Go
- Node.js
- Python
- Swift
- TypeScript
- React
- FastAPI
- LangGraph
Cloud & infra
- AWS Lambda
- Fargate
- Step Functions
- EventBridge
- DynamoDB
- OpenSearch
- CDK
- Terraform
GenAI & data
- AWS Bedrock
- Claude
- Multi-agent
- Semantic caching
- Vector hybrid search
- pgvector
- Qdrant
- LLM eval
Security & ops
- OIDC
- OAuth2
- RBAC
- SOC 2
- HIPAA
- HITRUST
- Secure SDLC
- OpenTelemetry
Writing
I write occasionally about the systems I build and the trade-offs I weigh, on Medium.
More long-form posts coming as I publish case studies for KloudArch and the Sage GenAI work.
Contact
Let's talk.
Not currently looking for job opportunities. Feel free to reach out for any architectural questions: hard infrastructure problems, multi-agent systems, RAG, evaluation, AWS.
- email reach.livin@gmail.com
- linkedin in/livin21
- github @livin21
- whatsapp +91 97441 42626