Software Engineer with 5 years of experience building large-scale systems serving 3M+ users. High-agency engineer with end-to-end ownership, from identifying opportunities and navigating stakeholder requirements to driving technical execution and deployment. Deep expertise in backend systems and Machine Learning infrastructure at scale.
Now Building the unified e-commerce stack at Kirana Club — OMS, logistics allocation, and a real-time ClickHouse pipeline doing 20M+ events/day.
Experience
Kirana Club Software Development Engineer Jan 2023 — Present
- Architected and led the development of a unified e-commerce platform (Go, Next.js) that orchestrates the complete order lifecycle for thousands of daily transactions, integrating a custom Order Management System (OMS) with automated logistics allocation and warehouse management for hundreds of sellers.
- Developed a centralized suite for SKU cataloging, automated pricing, and coupon management alongside a real-time analytics platform to track sales trends and supply chain bottlenecks, reducing order processing time from hours to near-instantaneous while significantly improving delivery success rates.
- Built an end-to-end customer support platform with designing the ticketing interface, agent workflow, and automated routing/prioritisation logic to manage thousands of daily inquiries. Streamlined processes across users, agents, and sellers to resolve exceptions, reducing TAT from hours to minutes and directly driving increased repeat orders.
- Engineered a high-performance spatial service using SQL spatial indexing and H3 Grid Indexing with bounding box optimizations, achieving <30ms latency for reverse geocoding and <100ms p99 for “Nearby Users” queries.
- Designed and deployed a self-managed ClickHouse cluster to handle 20M+ daily events for high-velocity impression data, optimizing data modeling and partitioning for high-throughput analytics.
Retail Pulse Machine Learning Engineer Aug 2020 — Jan 2023
- Architected and deployed a multi-stage product recognition pipeline for Indian retail environments combining YOLO object detector (trained on 25k+ images), DML embeddings (trained on 5M+ images) with FAISS-based matching, quality filtering, and screen recapture detection. The pipeline achieved 92% product-level and 85% SKU-level accuracy across 25,000+ SKUs, processing 10M+ images/month in production.
- Developed a client-facing analytics platform visualizing real-time store execution metrics, enabling major FMCG clients to monitor shelf compliance and stock availability.
Sparrosense Data Science Intern
- Developed self-supervised steel pouring classification in steel manufacturing process.
Humonics Global Data Science Intern Jun 2019 — Jan 2020
- Built instance segmentation models for automated car insurance claims achieving 0.8 mAP and optimized the inference pipeline using TorchScript and C++ to achieve a 1.5x speedup in production.
Technical Skills
Languages
Golang Python TypeScript SQL C++
Backend & Systems
FastAPI Node.js Next.js gRPC Docker RabbitMQ System Design
Data & Storage
PostgreSQL Redis ClickHouse Elasticsearch Firebase H3
Machine Learning
PyTorch YOLO OpenCV Deep Metric Learning Model Serving
Education
Guru Gobind Singh Indraprastha University · New Delhi, India B.Tech in Electronics and Communication Engineering · 2017 — 2021
Achievements
- Open Source Contributor: Contributed to OpenCV library with merged PRs.
- Scholarship: Recipient of Secure and Private AI Scholarship from Facebook and Udacity.
Recent Writing
adityak2920 / built with jekyll last updated · Jul 2026