Skip to content

Building AI Systems from Dubai

Shipping Production Software Since 2016

Ajay Singh
Parmar

Backend & AI Systems EngineerAI infrastructure · Model serving · Retrieval · Distributed systems

I build systems that survive production.

For the past decade I've engineered distributed backend platforms handling millions of requests. Today I'm applying those same engineering principles to production AI systems built around LLMs, retrieval, orchestration and reliable software.

ajay_singh_parmar_resume.pdf · updated 26 Jul 2026

Career-wide outcomes

5 metrics · career to date

Peak figures from systems I built — not current-role metrics.

10M+/ day

Ad auctions served

Audiomob RTB

~100ms

Response at p99

Audiomob RTB

2Kreq/s

Sustained throughput

Audiomob RTB

99.9%

Uptime at peak load

Audiomob

10+yrs

Shipping software

Career to date

Measured in production on the Audiomob RTB platform (Feb 2023 – May 2025), except where noted. Each number is attributed to the system it came from.

About

Dubai, United Arab Emirates

Ajay Singh Parmar
Based in
Dubai, United Arab Emirates
Shipping since
2016
Primary languages
Go · Python · TypeScript
Clouds
GCP · AWS
Domains
AdTech · Hospitality · Healthtech · Insurtech

Who I am

I am a software engineer who likes the problems that only show up under load.

Over the last decade — 10+ years now — I have designed distributed backend systems, cloud platforms and high-throughput services across AdTech, hospitality, healthcare and insurance, including payment flows where correctness is not negotiable.

Today my focus is production AI systems: retrieval over real transactional data, tool-using agent workflows, and the serving and orchestration layers that decide whether any of it holds up outside a demo.

What I build

AI systems
Retrieval over live transactional data, tool-calling agents, and keeping a model call from becoming the slowest hop in an otherwise fast request path.
Backend platforms
Multi-tenant services where reservations and money meet, with idempotent transaction handling and real consistency guarantees rather than hopeful ones.
Infrastructure
Terraform-managed GCP and AWS, Kubernetes, and enough instrumentation to know a system is degrading before a user tells me.

How I think

Languages, frameworks and models turn over every few years. The reasons systems fall over do not: unbounded queues, retries without backoff, state that two services both believe they own.

So I optimise for understanding systems deeply enough that the next generation of tooling is a detail rather than a re-education. That is why moving into AI infrastructure felt like continuity rather than a change of career — an inference call is just another dependency with a latency budget, a failure mode and a cost per request.

To discern the truth in everything, by whomsoever spoken, is wisdom.
Valluvar · Tirukkuṛaḷ 423, c. 5th century CE

Education

Jabalpur Engineering College

Bachelor's, Information Technology · GPA 8.0

July 2012 – June 2016

  • Committee member, Curiosity Club
  • Co-ordinator, Training & Placement Office

Outside work

  • Automation tools for daily productivity
  • Exploring cultures through travel
  • Photography
  • Bowling
  • Go-karting
  • Badminton
  • Cricket

Experience

6 roles · 2016 → present

  1. 2025 —

    KPTAC Technologies

    Senior Backend Engineer

    May 2025Present · Dubai, United Arab Emirates

    • Own end-to-end backend architecture for a multi-tenant hospitality SaaS platform (Digital eMenu), defining system design, API contracts, and data models across Orders, Payments, Reservations, Restaurant Management, Notifications, and CRM.
    • Architect scalable, fault-tolerant microservices handling high-volume restaurant transactions and real-time order flows, with concurrency control and strong consistency guarantees across reservation and payment systems.
    • Engineer secure payment integrations with idempotent transaction processing and webhook-based event handling, ensuring reliability under production load.
    • Integrate AI/LLM-driven features into the platform—menu recommendations, automated guest insights, and context-aware notifications—building RAG pipelines and LLM API integrations against transactional data.
    • Set technical direction for the team—driving architectural decisions, defining engineering standards, and mentoring engineers to raise delivery quality and velocity.
    • Partner directly with product, mobile, and frontend teams to translate business workflows into reliable backend systems, serving as the technical anchor across the stack.
  2. 2023 — 2025

    Audiomob Limited

    Senior Software Engineer

    February 2023May 2025 · Abu Dhabi, United Arab Emirates

    • Architected Audiomob's RTB platform from PoC to production, doubling throughput to 2K req/sec and sustaining ~100ms responses for 99% of 10M+ daily auctions through optimized GCP infrastructure (Bigtable, Redis) and custom pacing algorithms.
    • Built a gRPC service layer integrating ML models (CTR, LTR, CPI) for real-time ad targeting, cutting inference latency 30% and lifting ROI 25%.
    • Engineered low-latency infrastructure with Terraform, sustaining 99.9% uptime at peak load and re-architecting database schemas to halve query latency.
    • Redesigned the SSP Developer Dashboard (Svelte 5, Tailwind CSS) to cut user task time 40%, and automated BigQuery pipelines for real-time campaign tracking, reducing reporting lag 70%.
    • Owned the full system lifecycle—R&D, infrastructure, CI/CD, and cross-team alignment—ensuring platform performance met business KPIs.
  3. 2022 — 2023

    Akamai Technologies

    Senior Software Engineer

    March 2022January 2023 · Bangalore, KA, India

    • Engineered scalable REST API features for Router Configuration Management System (RCM) using Flask and React.js, handling configurations for 10k+ routers to streamline network operations.
    • Spearheaded backend database migration tool development for RCM, transitioning legacy systems to a dynamic environment and reducing configuration time by 30%.
    • Orchestrated full-stack workflows (GitHub/Jira/Jenkins) and collaborated with cross-functional teams to enhance network tools, cutting incident resolution time by 25%.
  4. 2021 — 2022

    Adcuratio Media Inc.

    Software Development Engineer II

    February 2021March 2022 · Bangalore, KA, India

    • Architected an event-driven Multi-Seller Platform using FastAPI with real-time notifications/logging (Kafka), enabling scalable transaction handling.
    • Led end-to-end development of core services (Subscriptions/Payments/Reviews) for Folksmedia Streaming (Python), mentoring juniors to boost team output.
    • Optimized API performance by 40% by applying SOLID principles and microservices architecture in cloud-native systems; designed DSP backend supporting addressable ads for agencies.
  5. 2019 — 2021

    Meddiff Technologies (Now Ramsoft)

    Software Engineer

    March 2019January 2021 · Bangalore, KA, India

    • Built a Quality Audit feature for the RIS/PACS server, increasing operational transparency and streamlining radiology workflows.
    • Pioneered a Tele-ultrasound web app (Python, Django, React), rapidly delivering remote ultrasound interoperability during COVID-19 to reduce in-person exposure risk, taking it from concept to working POC end to end.
    • Automated quality audits and server monitoring across a 50+ machine PACS fleet (Linux/Windows) and modernized the hospital inventory stack, cutting manual workflows 40% and enabling real-time radiology performance tracking.
  6. 2016 — 2019

    Infosys Limited

    Senior System Engineer

    December 2016March 2019 · Mangalore, KA, India

    • Implemented a critical policy-change feature for Universal Life and Whole Life insurance underwriting, enabling accurate handling of mid-term policy adjustments.
    • Restructured a core correspondence application with SOLID-principled Python backend code, improving performance and maintainability and boosting processing throughput by 20%.
    • Developed high-performance mainframe tools in REXX and PL/1 for CRUD operations on a DB2 database, improving reliability and reducing processing time.

Skills

6 groups · no self-rated scores

AI Engineering
LLM integrationRAGTool callingAgent workflowsContext engineeringVector databasesModel serving
Distributed Systems
gRPCKafkaPub/SubRedisMicroservicesEvent-driven architectureHigh-throughput APIs
Cloud & Infrastructure
GCPAWSKubernetesDockerTerraformCI/CD
Reliability
Idempotent transaction designWebhook event handlingConcurrency controlLatency budgetingLoad testing
Languages
GoPythonTypeScriptSQL
Data
PostgreSQLBigQueryBigtableRedisMongoDB

Also shipped, off the résumé skills line

  • Rust Tarjuman
  • Svelte 5 Audiomob dashboard
  • Vue.js CovidTracker
  • Flask Akamai RCM
  • REXX / PL/1 / DB2 Infosys

Selected work

2 of 4 shown

BuildingAIJul 2026

Tarjuman

Self-hosted, fully offline live speech transcription, translation and speaker diarization — in a single Rust binary.

  • Rust
  • TypeScript
  • Whisper
  • SQLite
  • WebSockets
  • Docker
Deployment
Single binary
Cloud APIs
None
Smallest target
Raspberry Pi

Engineering Notes

On Medium

Thoughts from building distributed systems, production AI software and backend platforms—covering the engineering decisions, tradeoffs and failure modes that rarely make it into tutorials.

Contact

Replies to my personal inbox

Interested in AI infrastructure, distributed systems or backend engineering? Whether you're hiring, collaborating or simply want to discuss architecture, I'd love to hear from you. For anything time-sensitive, WhatsApp is fastest.

Send a message

Elsewhere