SG — SYSTEM

SYSTEM ONLINE

SAHIL GHULE

Backend & AI Systems Engineer

I build systems that process data, power analytics, and turn complex workflows into reliable software.

EXPLORE SYSTEM
Location
Bengaluru, IN
Shipping
3+ years
Throughput
~1M events/day
Core stack
Python · FastAPI · ClickHouse
Status
AVAILABLE

01 / SYSTEM OVERVIEW

Engineer profile

Backend engineer with 3+ years designing and owning Python services and data-intensive platforms in production. I care about service boundaries that hold, query cost that is designed rather than discovered, and async paths that stay predictable when the volume arrives.

  1. Backend Engineering

    Owned

    Own the architecture and development of a distributed analytics platform end to end; built reusable API surfaces other teams consume.

    • Python
    • FastAPI
    • GraphQL
    • gRPC
    • asyncio

  2. Distributed & Async Systems

    Owned

    Re-architected async job execution from RQ to ARQ — 3× worker concurrency and 42% faster end-to-end processing.

    • ARQ
    • RQ
    • Redis
    • Scheduling
    • Concurrency

  3. Data Systems

    Owned

    Analytical schema design, aggregation strategy and query tuning over datasets ingesting roughly a million events a day.

    • ClickHouse
    • PostgreSQL
    • MongoDB
    • SQL
    • ETL

  4. AI / RAG

    Shipped

    Shipped a production RAG pipeline inside a live product; deployed ML pricing models serving real-time recommendations.

    • RAG
    • Parameter extraction
    • Random Forest
    • Training pipelines

  5. Production Engineering

    Applied

    Backfill millions of records into query-ready datasets without disturbing live workloads; profile and harden services under heavy load.

    • Backfills
    • Profiling
    • Reliability
    • Playwright
    • Cloud

02 / EXPERIENCE

Systems I have owned

Three roles, three systems. Each one is drawn below as the architecture it actually is — scroll a diagram to watch it assemble, or focus a node to read what it does.

  1. 02.1

    UrbanPiper

    Software Engineer II

    Aug 2025 — Present · Bengaluru, India

    System Periscope — a distributed availability & downtime analytics platform for enterprise food-delivery operators.

    • ~1Mevents / day
    • 10,000+stores monitored
    • 3×worker concurrency
    • 42%faster job processing
    • Own the architecture and development of Periscope, a distributed backend platform on Python, FastAPI, ClickHouse and asynchronous workers, serving availability and downtime analytics to enterprise customers.
    • Design high-throughput ingestion and analytics pipelines processing ~1M events per day across 10,000+ stores — 15-minute polling, 96 checks per store daily.
    • Re-architect asynchronous job execution by migrating from RQ to ARQ, increasing worker concurrency 3× and reducing end-to-end job processing time by 42%.
    • Build ETL and backfill systems that process millions of records into query-ready analytical datasets without disrupting live workloads.
    • Build large-scale scheduling and automation workflows integrating multiple external systems, reducing manual operational effort by 70%.
    • Develop distributed crawling and indexing pipelines with Playwright for automated real-world data acquisition.
    • Model and optimise ClickHouse schemas and SQL queries over large distributed datasets to keep analytical reads fast as data volume grows.
    • Python
    • FastAPI
    • ClickHouse
    • ARQ
    • Redis
    • Playwright
    • asyncio
    • SQL
    Architecture — Periscope Hover or focus a node to inspect · Enter to pin

  2. 02.2

    Apperture AI

    Software Engineer

    May 2024 — Aug 2025 · Remote

    System Availability intelligence & applied GenAI — analytical dashboards over food-delivery availability data, plus a production RAG pipeline inside a live product.

    • 70%less dataset prep time
    • RAGshipped to production
    • 2reusable API surfaces
    • Integrated parameter-extraction models into a production Retrieval-Augmented Generation pipeline, shipping an AI-powered capability inside a live product.
    • Automated AI training-data generation pipelines with reproducible Python workflows, reducing dataset preparation time by 70%.
    • Built backend services powering analytical dashboards that monitor store- and item-level availability across food-delivery ecosystems.
    • Designed scalable analytical storage and query systems on ClickHouse — schema design, aggregation strategy and query tuning — to sustain large analytical workloads.
    • Built reusable GraphQL and gRPC services for efficient large-scale data access, and optimised backend services for performance and reliability under heavy load.
    • Python
    • GraphQL
    • gRPC
    • ClickHouse
    • MongoDB
    • RAG
    • asyncio
    Architecture — Analytics & RAG Hover or focus a node to inspect · Enter to pin

  3. 02.3

    Cogoport

    Software Development Engineer I

    Feb 2023 — Oct 2023 · Mumbai, India

    System Dynamic pricing for logistics — machine-learning pricing models deployed to production for real-time recommendations on large-scale logistics data.

    • 15%better pricing accuracy
    • Livemodels in production
    • Developed and deployed machine-learning dynamic pricing models (Random Forest, Python) to production for real-time pricing recommendations on large-scale logistics data.
    • Built backend services powering logistics analytics dashboards, improving pricing prediction accuracy by 15% through data-quality improvements and feature engineering.
    • Python
    • Random Forest
    • Scikit-learn
    • Pandas
    • SQL
    Architecture — Pricing pipeline Hover or focus a node to inspect · Enter to pin

03 / SYSTEMS MAP

The stack, as a dependency graph

Technologies are not a list of badges — they are a graph of things that call each other. Hover a node to isolate what it touches and what it made possible.

SELECT A NODE

  • Python
  • SQL
  • C++
  • FastAPI
  • GraphQL
  • gRPC
  • ClickHouse
  • PostgreSQL
  • MongoDB
  • ARQ
  • Playwright
  • Pandas
  • Scikit-learn

04 / CASE FILES

Selected work

Personal systems and studies, filed as cases. Open one for the problem, the approach, the architecture and what actually came out of it.

  • PRJ_001 SHIPPED

    Applied deep learning · Medical prognosis

    A deep-learning application predicting prostate cancer probability, served through a Flask interface and deployed to cloud infrastructure.

    • Python
    • Flask
    • DNN
    • Cloud
    REPO ↗
  • PRJ_002 RESEARCH

    Regression modelling · Operational analytics

    Machine-learning models predicting full-load electrical power generation, aimed at operational-efficiency decisions.

    • Python
    • Scikit-learn
    • Pandas
  • PRJ_003 EXPLORATION

    Decentralised media · Web3

    A short-form video client exploring decentralised media patterns — a TypeScript frontend over blockchain integration experiments.

    • TypeScript
    • Blockchain
    • Web3
    REPO ↗
  • PRJ_004 RESEARCH

    Exploratory analysis · Visualisation

    Notebook-driven analytics over real datasets — exploration, visualisation and insight generation.

    • Python
    • Pandas
    • SQL
    • Jupyter
    REPO ↗

05 / RESEARCH

Publication

Peer reviewed

Prostate Cancer Prognosis — A comparative approach using Machine Learning Techniques

5th International Conference on Intelligent Computing and Control Systems (ICICCS) · December 2020

Abstract

A comparative evaluation of machine-learning techniques applied to prostate cancer prognosis, benchmarking classifier families against one another on the same task to establish which approaches carry genuine predictive signal.

Topics

  • Machine Learning
  • Comparative Study
  • Medical Prognosis
  • Classification

Related

The applied counterpart to this study shipped as PRJ_001 — AI-Powered Prostate Cancer Detection.

06 / CONNECTION TERMINAL

Have a system worth building?

connection_terminal — sahil.ghule STATUS: AVAILABLE

open --channels

4 channels open · Bengaluru, IN · replies within a day

INITIATE CONNECTION

Open to backend platforms, analytics infrastructure and AI systems in production.