Harsh Kochar

Software Engineer
Bengaluru, IN.

About

Highly proactive Backend Software Engineer with over 3 years of experience in a leading fintech startup, specializing in Java, Python, Go, Rust, and JavaScript. Driven by a strong interest in problem-solving, I leverage artificial intelligence for automation and process optimization, consistently delivering impactful, user-centric solutions. Proven ability to take ownership, navigate complex technical challenges, manage stakeholders effectively, and foster best practices. Thrive in dynamic startup environments, focusing on identifying product use cases and delivering high-impact solutions.

Work

PhonePe
|

Software Engineer

Summary

Contributing to a platform team at India's largest fintech startup, focusing on building scalable and resilient backend systems and driving engineering excellence.

Highlights

Engineered a real-time batch rule evaluation system processing over 22,000 queries per second (QPS) with 99.98% uptime and a P99 latency of 100ms, effectively preventing fraudulent transactions and saving millions of rupees monthly.

Designed and implemented asynchronous orchestration for periodic rule evaluations using Spark, Airflow, and Kafka event streams, optimizing accuracy and performance by reducing live service load bursts by 17%.

Enabled critical business use cases, including trade validations, compliance checks, business validation, and Anti-Money Laundering (AML) fraud monitoring.

Developed a tooling platform with an automation framework for schema generation and migration, resulting in a 90% reduction in build time and a 70% increase in onboarding speed.

Engineered a scalable Experimentation Suite for rapid iteration of rules and ML models, incorporating real-time confusion matrix, controlled rollouts, and A/B testing, which reduced false positives by 32%, false negatives by 5%, and runtime errors by 37%.

Architected an AI-powered interview feedback tool utilizing Gen-AI for transcription and metric-based discussion tagging, reducing review feedback time by 80%.

Led a comprehensive revamp of failure handling mechanisms, unifying logging, metric ingestion, and error management, which decreased upstream errors by 7% and reduced data ingestion by 20% through improved code structure and design patterns.

Implemented robust failure handling, including fallback mechanisms for data operations, auto-rollback for system failures, and circuit breakers, ensuring high system stability and data integrity.

Standardized and consolidated critical workflows, including test rule archival, Kafka event schema generation, data backfill, production mirroring, and analytical pipelines for OLAP DBs (sync and async).

Spearheaded critical application upgrades, including migrating from Java 8 to 17, significantly improving maintainability by reducing JAR sizes through dependency cleanups and API deprecation.

Restructured and unified Data Access Object (DAO) and database interaction layers for Aerospike, establishing new team-wide architectural patterns.

Guided the design and implementation of rule failure alerting, integrating with monitoring tools to detect abnormal spikes and block rates, and re-modeled API layers for broader service adoption.

Led the design for multi-vendor investigation, transitioning from a reactive human-driven approach to an automated, scheduled system, optimizing cost and efficiency in data sourcing.

Designed a comprehensive testing suite with CSV upload support and real-time context enrichment capabilities via data mutators.

Developed a robust storage management strategy for efficient handling of result information and downstream triggers.

Provided technical leadership, leading disaster recovery (DR) drills, overseeing database migration activities, mentoring team members, and conducting interviews for platform engineering roles.

PhonePe
|

Engineering Development Intern

Summary

Contributed to backend development and data engineering initiatives within a leading fintech platform.

Highlights

Optimized rule performance by developing confusion matrix methods for anomaly detection and leveraging Hive for big data queries; implemented Directed Acyclic Graphs (DAGs) on Aerospike for efficient data procurement.

Applied optimized rule performance methods for A/B testing and traffic shaping within the Experimentation platform.

PhonePe
|

Engineering Development Intern

Summary

Focused on data pipeline development and clustering for location-based services.

Highlights

Devised an ETL pipeline using Spark for generating Top K locations and implementing geofencing capabilities.

Reused multi-density clustering techniques to enhance location-based insights.

Education

Indian Institute of Information Technology, Allahabad

Bachelor of Technology

Information Technology

Courses

Artificial Intelligence and Automation (Minor)

Data Structures and Algorithms

Object-Oriented Programming

Database Management Systems

Operating Systems

Machine Learning

Artificial Intelligence

Deep Learning

Reinforcement Learning

System Design

Awards

PhonePe-group Hackathon Finalist

Awarded By

PhonePe

Achieved finalist standing for developing an English to Multi-DSL converter using ChatGPT APIs.

PhonePe Bootcamp Hackathon - 1st Place

Awarded By

PhonePe

Secured 1st place for developing an End-to-End Functional Shopping Platform.

Student Welfare Council Secretary

Awarded By

SWC Gymkhana

Served as Student Welfare Council Secretary, contributing to student welfare initiatives.

IIITA B-Hack - Most Innovative Idea

Awarded By

IIITA

Recognized for the most innovative idea in Track-2 Business.

New Gen IEDC - 3rd Place

Awarded By

New Gen IEDC

Awarded 3rd place for the Instagram Bot Detector project.

Inter-year Football Championship - 2nd Place

Awarded By

IIT Allahabad

Achieved 2nd place in the Inter-year Football Championship in both 2018 and 2019.

JEE Mains - 99.71 Percentile

Awarded By

Joint Entrance Examination (JEE)

Achieved 99.71 percentile in the Joint Entrance Examination (JEE) Mains.

JEE Advanced - 99.57 Percentile

Awarded By

Joint Entrance Examination (JEE)

Achieved 99.57 percentile in the Joint Entrance Examination (JEE) Advanced.

Publications

Demand response optimization for Microgrid Clusters

Published by

IIT KGP (Presented at 12th ICCCNT)

Summary

Co-authored and presented research on 'Demand response optimization for Microgrid Clusters' at the 12th ICCCNT, coordinating a team of 5. The work involved simulating a dynamic environment with an LSTM Model and implementing various Deep Reinforcement Learning models (Policy-based, Reinforce Agent, Value-Policy Hybrid, Asynchronous Advantageous Actor-Critic (A3C)). Successfully ran multiple models in parallel on a local CUDA setup, identifying and resolving PyTorch multiprocessing limitations.

Languages

English

Fluent

Hindi

Native

Skills

Tools & DevOps

Poetry, pip.

Programming Languages & Frameworks

Java, Core Java, Python, Shell, Dropwizard, Service Discovery, Scala, asyncio, multiprocessing.

Programming Languages & Frameworks

Go (golang), go-routines, Node.js, JavaScript, C++, Django, Flask, FastAPI.

Programming Languages & Frameworks

LaTeX, Prolog, Matlab, Lua, Pillow.

Data Technologies

Aerospike, Spark, Kafka, Hive, Airflow, Redis, MySQL, PostgreSQL, MongoDB.

Data Technologies

Avro, MariaDB, Elastic Search, Numpy, Hadoop, Caffeine Cache, MVEL.

Data Technologies

GraphDB, HBase, ProtoBuf.

ML/AI & Data Science

Machine Learning, Deep Learning, Artificial Intelligence, Reinforcement Learning, Data Mining, Anomaly Detection, Clustering, Analytics.

ML/AI & Data Science

PyTorch, TensorFlow, CUDA.

ML/AI & Data Science

Hugging Face, Ollama.

Tools & DevOps

UNIX, Git, Jenkins, Docker, Apache Maven, Jira, CI/CD, Grafana.

Tools & DevOps

Marathon, Mesos, A/B Testing, Obsidian, vim, Ops-genie, Confluence, Prometheus, AWS.

Core Concepts & Methodologies

Algorithms, Data Structures, System Design, Performance Optimization, Asynchronous Programming, Multiprocessing, Design Patterns, Big Data Management and Analytics, Maintainability, Data Modeling, Data Warehousing, Problem-Solving, Backend, Performance Tuning, Software Development.

Core Concepts & Methodologies

Research, Convex Optimization.

Leadership & Product

Leadership, Mentoring, Product Strategy, Process Improvement, Stakeholder Management.

Projects

MuDi Stream (Multi-density Clustering for Live Data Stream)

Summary

A project focused on classifying random datastreams using multi-density clustering.

Instagram Bot Detector (Random Forest Classifier)

Summary

Developed a random forest classifier to distinguish farming bots from human users.