i.
Don't romanticize code.
The real job is breaking down problems, spotting patterns, and delivering the right solution — code or otherwise.
Ankit Raj.
Builds systems that ship, survive, and scale.
6+ years building distributed systems across Procurement, Receivables & CRM.
01 — About
The short version.
Seasoned backend with diverse experience across SaaS teams like Jaggaer, LeadSquared, and HighRadius. I specialise in building systems that handle messy data, third-party integrations, event-driven pipelines, and legacy system migrations.
I bring a strong product mindset and solid proficiency in cloud, devops, and AI, enabling me to collaborate effectively and deliver end-to-end solutions.
years of experience in building production-grade systems
Philosophy
ii.
Readable code is non-negotiable.
Intent-revealing code, structured design and documentation are backbone of fast and efficient team collaboration.
iii.
Engineering is team sport.
Clear articulation, communicating, negotiating, and delegation will always outperform individual heroics.
Skills
Backend
Data & Messaging
Cloud & Infra
Architecture
02 — Experience
Where I've shipped.
What I built
- Re-architected third-party integrations (Carbmee, Trustpair, Veem) into a dedicated microservice — phased rollout with zero downtime.
- Built an event-driven datasync pipeline streaming procurement data (requisitions, bidding, auction events) from a legacy system to the new platform at 2,000+ events/sec.
- Shipped an admin console + real-time insights dashboard sitting behind a unified gateway handling 1,000+ RPS.
Architecture & impact
- Authored a shared resilience library — retry, rate-limiter, circuit-breaker, fallback, DLQ, and backpressure — built on bucket4j, resilience4j, and RabbitMQ. Reused across services.
- 100% reduction in third-party-integration production incidents after the migration.
- Decoupled integration logic from core business workflows, restoring readability and aligning the codebase to SOLID principles.
Tech
- Java, Spring Boot, JPA, GraphQL, MySQL
- RabbitMQ, ExecutorService, resilience4j, bucket4j
- JUnit, Mockito, Docker, Argo CD
What I built
- An email service microservice integrated with SendGrid, SES, Gmail, and Outlook — 10K+ emails/day with an automated integration API.
- A multi-channel notification service (Email, SMS, WhatsApp) built on the factory pattern, with monitoring and automated retries — 50K+ notifications/day.
- An email metrics portal + SOPs that protect shared IP reputation by quarantining risky client activity.
Architecture & impact
- Layered rate limiting (Bucket4j), AI-driven spam detection (AWS Bedrock), Redis-backed draft caching, and SQS + EventBridge retries to deliver high-throughput concurrency.
- Standardised trunk-based development with weekly releases and rollback playbooks — faster, safer deploys.
- Led a cross-region environment migration with the DevOps team — zero downtime. Mentored 3 interns.
Tech
- Java, Spring Boot, Bucket4j, Micrometer
- AWS SES, SQS, EventBridge, Bedrock, EC2
- Redis, MySQL, Jenkins, Argo CD, Python
What I built
- A payment file processor & parser handling 1B+ records/year, paired with a concurrent rule engine built on ExecutorService.
- An intelligent L3 credit-card payment flow and tightened inter-service integration that cut payment failures by 10–12%.
- A correspondence configuration management service (CRUD) handling 1M+ records/year.
Product ownership
- Owned an internal utility product (1,000+ DAU) end-to-end — stakeholder discovery, roadmap, MVP, and adoption.
- +10% bandwidth utilisation and +5% implementation speed across the delivery org.
- Drove adoption by tracking dashboard CTR and iterating on the gaps it surfaced.
Tech
- Java, Spring, Hibernate, ExecutorService
- MySQL, Redis, REST
- Bash, Git, Jenkins
03 — Selected Work
Featured projects.
Inbox Intelligence
Personal project · In progress
The problem
- Personal inboxes drown in job alerts, banking, OTPs, promotions, and newsletters. Built-in labels are static and don't learn from how you actually triage.
- I wanted a system that could read, group, and label emails by meaning — not by sender or subject keywords — and adapt as I gave it feedback.
What I'm building
- A Spring Boot pipeline that pulls emails over the Gmail API, sanitises HTML/markup down to clean text, and persists the result alongside its embedding vector.
- An embedding + clustering layer using Ollama
(
nomic-embed-text) for vectors, pgvector for storage, and cosine similarity + DBSCAN for auto-grouping. - An LLM labelling stage that names each cluster from a handful of representative samples, then assigns the label back to every member email.
- A feedback loop: user actions (move, mark important, mark spam) feed back into cluster boundaries and label quality over time.
Architecture thinking
- Pipeline, not monolith.
Clean → Store → Embed → Similarity → Cluster → Label → Feedback— each stage is its own service boundary so I can swap models or storage independently. - Local-first by default. Ollama runs the embedding model on-device so no email content leaves the machine. Cloud LLMs are an opt-in for the labelling stage only.
- Embeddings are the source of truth. Clustering and labels are derived state — they can be rebuilt from vectors, so experimenting with new clustering algorithms is cheap.
- Idempotent ingestion. Gmail's history-id watermark plus deterministic content hashes mean re-runs and retries never double-process an email.
Architecture diagrams and a deeper walkthrough are coming soon.
Product Manager Fellowship
NextLeap · 2023
Increasing Reviews on Zomato
- User research with 100+ respondents showed 50% of frequent users lacked incentive to write reviews.
- Designed a gamification system with redeemable virtual coins targeting 2.1M users, with spam-mitigation built in.
- Wireframes, system architecture, and review-to-order ratio as the success metric, with a long-term scaling plan.
Subscriber Growth for YouTube Music
- Research with 200+ respondents identified missing equalizer/crossfade as a blocker for 33% of upgrades and a retention risk for 40% of premium users.
- Designed an AI-powered equalizer that adapts to listening history.
- Wireframes, architecture, completion-rate metric, and a plan for the cold-start problem.
DevOps Certification
GeeksForGeeks · Expected 2026
Automated CI/CD Pipeline for Portfolio Deployment
- GitHub Actions pipeline for automated portfolio deploys.
- S3 for storage, CloudFront for CDN + cache invalidation.
- Route 53 for DNS, ACM for TLS — secure by default.
04 — Education
Where it started.
Bachelor of Technology · Computer Science & Engineering
Kalinga Institute Of Industrial Technology
05 — Contact
Let's build something good.
Hiring for a backend role, scoping a hard distributed-systems problem, or just want to compare notes on resilience patterns? My inbox is open.
rja.ankit@gmail.com