Headline

Backend & ML Engineer @ Guidewire | Java · Python · Apache Spark · Iceberg · AWS · LLMs | Building event-driven data pipelines at scale

About

Backend and ML engineer with 4+ years building data-intensive systems — from event-driven ingestion pipelines on AWS to neural-network recommendation models that moved real product metrics. Currently at Guidewire, I architect and lead IP3 — the Iceberg-tier sink in our 3-stage CDC pipeline (DB WAL → Kafka/Debezium → Iceberg) — serving 400+ DP-IDs across 100+ tenants and 700+ tables. Recent shipping highlight: replaced 40 per-DP-ID Lambdas with a unified Spark job using a single Iceberg commit, eliminating commit contention and cutting data-gap detection runtime ~20x. Also reduced an Iceberg cleanup job from a 48+ hour timeout to under 40 minutes by resolving race conditions with concurrent writers. Previously at CaaStle, I worked across the AI/ML stack — built a three-tower neural network recommendation model (10% lift in closeting rate), embedding-based product recommendations using LangChain and GPT-4V (5% engagement lift), and AI-generated fashion content via LoRA fine-tuning on Stable Diffusion. Also led the Collection Generation pipeline, cutting runtime from 5 hours to 30 minutes. Before that, I built secure React Native flows for PIN reset and money add/withdrawal at Groww, used by 70M+ users. What I care about: diagnosing the architectural root cause (commit contention, race conditions, throttling logic) rather than throwing more compute at the symptom. I enjoy taking ambiguous infra problems from POC through to team-wide adoption. Stack I work in: Python · Java · Apache Spark · Apache Iceberg · Kafka · AWS (Lambda, EventBridge, Glue, Athena, EMR Serverless, SageMaker) · Elasticsearch / ELK · LangChain · PyTorch · Spring Boot. Open to senior backend, ML, and data-platform roles. Bangalore or remote.

Location

Bengaluru

AI stack

claude code

gemini

github copilot

openai

Work experience

Software Engineer 2 at Guidewire Software, Jul 2025 - Present · 1 yr 2 mos, Bangalore Urban, Karnataka, India, Building and operating Guidewire's data ingestion platform — an event-driven pipeline that lands changes from upstream insurance databases into queryable Iceberg tables across 400+ data partitions and 100+ tenants. My work spans architecture (designed and led IP3, our Iceberg-tier sink), reliability (DB rollback handling, per-DP-ID alerting redesign), performance (parallel Spark for data-gap detection — ~20x faster than the prior Lambda fan-out), and security (resolved 50+ Brinqa vulnerabilities across 3-5 services).

Senior Software Engineer at CaaStle, Dec 2024 - May 2025 · 6 mos, Bengaluru, Karnataka, India, AI/ML and backend engineering for The Closet, CaaStle's subscription clothing rental platform. Built recommendation systems (three-tower neural network, embedding-based via LangChain + GPT-4V), AI-generated fashion content (LoRA fine-tuning + Stable Diffusion on AWS SageMaker), and Java/Spark data pipelines that powered personalized collections. Owned the Collection Generation pipeline end-to-end and developed multiple Spring Boot microservices.

Data Driven Applications Engineer at CaaStle, Jul 2022 - Jan 2025 · 2 yrs 7 mos, Bengaluru, Karnataka, India

Software Application Developer at Groww, India, Jan 2021 - Jul 2021 · 7 mos, Bengaluru, Karnataka, India, Consumer fintech mobile engineering trainee. Built secure React Native flows for sensitive money operations (PIN reset, add/withdraw funds) used by 70M+ users, plus reusable UI components for the broader app.

Education

Bachelor of Technology - BTech, Computer Science, Punjab Engineering College

Public links

github: https://github.com/Akash-Kait

leetcode: https://leetcode.com/akashkait