Skip to content

Instantly share code, notes, and snippets.

@ikouchiha47
Last active June 8, 2026 08:23
Show Gist options
  • Select an option

  • Save ikouchiha47/e0049a117c1d31838e5a25b11dc14cfe to your computer and use it in GitHub Desktop.

Select an option

Save ikouchiha47/e0049a117c1d31838e5a25b11dc14cfe to your computer and use it in GitHub Desktop.
Reume Typist

Amitava Ghosh

Staff Software Engineer — India


Profile

I'm a platform engineer who enjoys building scalable, fault-tolerant systems and making them faster and more reliable. I spend a lot of time profiling and benchmarking to uncover bottlenecks, reduce production issues, and improve system stability. I've led architectural decisions through ADRs, supported teams in building services, and worked across teams to improve delivery. Working across different parts of the platform, and my familiarity with some low-level debugging tools has also made me comfortable adapting quickly to new domains and problems.


Skills

Category Details
Languages Go, Javascript, Rails, NodeJS, Lua, Python, Bash
Databases MySQL, PostgreSQL, Redis, DynamoDB*
Technologies AWS, Terraform, Docker & Containers, gRPC, Protobuf, Redis Streams, SDK & DSL Design
Applied AI Dspy, CrewAI, Langchain, Langtrace, Prompt and Context Engineering, Workflow Orchestration, Embeddings and Applications
Other Skills Distributed Systems, Overservability, Few Unix tools, Zig

Experience

Scalarity — Bengaluru, India

Founding Engineer · 12/2025 - 02/2026

  • Started consulting full time with the previous CTO to build the initial platform, detangling what to build vs. what researchers actually needed — translating ambiguous research workflows into a concrete product surface.
  • Embedded with scientists and researchers to understand how they read papers, run experiments, and where they want acceleration; built working familiarity with the materials science domain (DFT, GPAW, matgen, figshare datasets) to make informed architectural calls.
  • Sketched the early platform architecture around data federation and inter-institution data movement (universities, labs, public datasets), so experiments and provenance could flow across boundaries without centralizing every dataset.
  • Deployed multiple ML frameworks needed for DFT calculations and predicting downstream features using GPAW, wiring them into a reproducible pipeline.
  • Compressed the materials data available on figshare (from 2GB to 300Mb), reducing storage and transfer cost for next-gen materials projects.
  • Built the PDF reader to convert papers into a hierarchical structure and improved semantic search using different embedding models, generating candidate questions for researchers to explore.
  • Built the Experiment agent and a validation ecosystem that ingests historical research papers and tries to produce a final paper, while catching low-quality or unsound ones.

Affogato (formerly Rendernet) — Bengaluru, India

Senior Software Engineer · 12/2024 - 12/2025

Desigining systems to support media generation and delivery, evolving schemas, MVPs and POCs to help product make decisions.

Architecture & System Design Ownership

  • Designed an orthogonal architecture for a collaborative video editor, cleanly separating timelines, assets, workflows, and rendering concerns to enable independent evolution and rapid experimentation.
  • Architected a workflow DSL backed by graph execution with concurrency controls, enabling faster integration of media-generation pipelines and reducing coordination overhead.
  • Designed fractional indexing–based primitives for non-linear collaborative editing, avoiding global reindexing and write contention.
  • Defined asynchronous, event-driven system patterns with queues and notifications to decouple user actions from long-running media processing and external dependencies.

Platform Engineering

  • Owned platform-level architectural decisions across media generation, search, authorization, moderation, and gateway services, balancing latency, cost, and operational complexity.
  • Designed a team-based authorization system and centralized reverse-proxy gateway with rate control and degraded-mode handling for external APIs.
  • Built internal DSLs for cross-embedding, full-text search, and metadata enrichment to standardize retrieval and evaluation across products.
  • Owned internal developer platform decisions, including CI/CD standards, QA gates, and automated PRD→ADR workflows using Claude-Code / Opencode agents.
  • Standardized deployment, configuration, and review workflows across services, reducing cognitive load and improving developer velocity.

Image & Video Optimization

  • Architected a dynamic image delivery pipeline (Golang + libvips) with on-the-fly and background compression, and dynamic go worker scaling, reducing transfer for 1K+ images from 4GB to 10MB.
  • Drove upstream libvips and govips improvements by shifting to reader-based APIs, enabling streaming processing and automated C-bindings.
  • Designing a scalable ffmpeg farm for video compression, delivery, and editing workflows (and AWS MediaConvert for later).

Agentic AI & LLM Infrastructure

  • Build agentic AI topologies, combining tools for faster scaffolding; optimized LLM Router with shorter prompts + vector search; developed internal tooling for rapid prompt experimentation.
  • Visual Intelligence: Built video analysis platform to extract screenplay structure, colors, and product metadata—foundation for knowledgebase refinement and campaign evaluation.
  • Building a critique system, combining LLM based scoring, embedding scoring, readibility scores, emotional scores. (Learning about Natural Language)

Development Velocity

  • Set up CI/CD pipelines, QA checks, and Claude-Code/Opencode powered agents to accelerate PRD->ADR conversions, code reviews, tests, and refactors.
  • Bridged the gap between product team and dev team for faster iteration on prompt changes.
  • Other smaller convention changes, to match the deployment readiness of frontend-backend releases, reducing late nights and multiple refactors post release.

TheBackendCompany — Bengaluru, India

Freelance & Consulting · 12/2023 - 12/2024

  • Designed and built a web-based CSV processing system using SQLite WASM for fast, local analysis and transformation of large datasets.
  • Consulted startups on early-stage architecture, infrastructure cost modeling, and cloud adoption strategies.
  • Built github.com/go-batteries, a reusable platform toolkit for scaffolding backend services and internal infrastructure.
  • Built LLM-backed APIs using Ollama and LangChain for text categorization and sentiment analysis experiments.
  • Founded and operated a small D2C apparel brand, owning end-to-end execution across product design, vendor sourcing, tech, and operations.

Sequoia Group — Bengaluru, India

Senior Software Engineer · 08/2020 - 07/2023

Tech Stack: Golang, Mysql, Redis, AWS, DynamoDB, Python, Airflow, Apache Benchmark, Linux tools: gdb, valgrind, perf.

  • Led the development of multiple new features and addressed SSO login issues, centralizing CSV management across use cases for clients like Snap and Mongo, and handling file encoding issues.
  • Build the developer platform for engineers for faster development, including internal SDK, docs generation, database migrations, PII log filters, merge conflict predictor etc.
  • Orchestrated the migration of unencrypted files and integrated Datadog APM to address slow queries and later introducing application and system metrics, across multiple teams.
  • Led efforts for Securing APIs with rate-limiting and more stringent request validations in collaboration with Security Team.
  • Collaborated on building an OpenID-based OAuth2 layer for employment history sharing, while resolving database locking issues from syncing new users and handling login requests, and maintaining compliance.
  • Developed and maintained core services, including a notification system using AWS (SQS, SES, SNS) and an internal developer platform to streamline workflows (PII log filters, database migrations, and more).
  • Standardized processes through the Golang SDK for configuration management, request tracing, and file encryption, while debugging production issues and resolving Redis connection issue to enable token caching and rate limiting.
  • Other notable work includes building a DB resolver library for Gorm V1, migrating code from Lambda to Airflow, and improving performance through pprof benchmarking and Flamegraphs.
  • Conducted code reviews, lectured on Go concurrency and Node.js Promises, and assisted teams in integrating with existing services, including generating Swagger docs.

Gojek — Bengaluru, India

Product Engineer · 2016 - 2020

Tech Stack: Golang, Rails, Postgres, Redis & Twemproxy, Ansible, Wireshark, Lua, Kong(nginx).

  • During the monolith decomposition of customer service, I worked on OAuth-based authentication and authorization, otp sending, rebalancing with multiple provider, migrating from Rails to Golang, and setting application and system metrics.
  • Led the customer gateway migration from HAproxy to Kong, improving API security, reducing AWS costs, and load optimization. This involved learning Lua, writing authentication plugins, setting system limits, planning percentage rollouts, and analyzing tcpdumps to resolve reconnection issues.
  • Improved API whitelisting and fraud prevention, providing the security team with better visibility over public-facing APIs.
  • Helped out security team, as they now had greater visibility and control over public facing API whitelisting, preventing frauds.
  • Other Contributions:
    • Implemented SymSpell to enhance ElasticSearch's Global Search service, improving fuzzy search accuracy and cross-selling, based on alpha-user feedback.
    • Helped resolve Kafka failures during GCP live migration by adjusting system parameters (ZOO_TICK_TIME, migration policies, and API timeouts) to prevent stream leader crashes.
    • Contributed to the internationalization phase by adding location headers and localizing error messages in collaboration with translators to translate app errors.
    • Played a key role in Gojek's e-commerce project, managing inventory synchronization and coordinating driver bookings across multiple services using existing APIs.
    • Build and maintained a slack bot to get boxes and other system information, translations.

Leftshift — Pune, India

Software Engineer · 08/2016 - 09/2016

Tech Stack: Nodejs, Javascript, React, Electron, Rails.

  • Worked on a project to periodically get Gojek app reviews, writing API and database structure, for data team to analyze, etc. (Acquired by Gojek)

Kreeti Technologies — Kolkata, India

Software Engineer · 2014 - 2016

Tech Stack: Rails, Postgres/Mysql, Redis, React, Nodejs, Phoenix(Elixir).

  • Setup of the full stack of Memento, a social network app, database design, in MERN stack and deployment with Heroku.
  • Implementation of login and account creation for CPG and Retailer, authentication, and authorization in Rails. Worked with UI/UX engineers. Photo upload using s3 and ImageMagick for image formatting.
  • Implementing a commenting feature on the promotions page for communication between multiple parties, Email-based weekly and daily reminder of promotions with ActiveJob, sidekiq.
  • Rebuilding the internal payslip generation, with dashboard in Elixir.
  • Matrimony app, handling caste system with multitenancy, handling authentication with devise, tenancy with act_as_tenant, search and angular frontend.

Education

West Bengal University of Technology — Kolkata, India B.Tech in Electronics & Communication Engineering · 2010 - 2014

#import "@preview/modern-cv:0.10.0": *
#show link: underline
#show: resume.with(
author: (
firstname: "Amitava",
lastname: "Ghosh",
email: "amitava.dev@proton.me",
phone: "(+91) 629-409-7693",
github: "ikouchiha47",
homepage: "https://ikouchiha47.github.io",
linkedin: "segfault-survivor",
address: "India",
positions: (
"Staff Software Engineer",
"Building Scalable, Fault-Tolerant Distributed Systems",
),
),
date: datetime.today().display(),
language: "en",
profile-picture: none,
colored-headers: true,
show-footer: false,
)
// Override template font size globally (must come AFTER the show rule)
#set text(size: 10pt)
#set par(leading: 0.5em)
= Profile Summary
Platform and Distributed Systems Architect with extensive experience designing and scaling mission-critical, high-throughput backend systems. I specialize in bridging the gap between ambiguous scientific/business requirements and concrete, reliable product surfaces. Proven ability to lead architectural decisions (ADRs), build internal developer tooling (DevEx), and execute complex performance and reliability improvements across entire service ecosystems. Deep technical fluency in low-level debugging, cloud-native patterns, and modern AI/LLM infrastructure.
\
// Shrink everything from here down (Technical Skills + Professional Experience)
#set text(size: 6pt)
= Technical Skills
#resume-skill-item(
"Languages",
(strong("Go"), strong("JavaScript"), strong("Rails"), "NodeJS", "Python", "Bash", "Lua", "Zig"),
)
#resume-skill-item(
"Data & Databases",
(strong("PostgreSQL"), strong("MySQL"), "DynamoDB", strong("Redis"), "SQL"),
)
#resume-skill-item(
"Platform & Tools",
(strong("AWS"), "Datadog", "TICK stack", "AppSec", "CI/CD", "Queues", "Linux Tools", "Nginx", "Terraform", "SDK/DSL Design", "Developer Experience"),
)
#resume-skill-item(
"AI & ML Infrastructure",
("Dspy", "Langchain", "CrewAI", "Langtrace", "Prompt/Context Engineering", "Workflow Orchestration", "Embeddings"),
)
= Professional Experience
\
#resume-entry(
title: "Scalarity",
location: "Bengaluru, India",
date: "12/2025 - 02/2026",
description: "Founding Engineer",
)
\
#resume-item[
_Consulting with previous CTO to build a scientific data platform_
- *Scientific Platform Architecture:* Defined and materialized the core platform architecture around data federation and inter-institution data movement, allowing experimental provenance and data to flow seamlessly across academic boundaries without centralizing massive datasets.
- *Domain Translation:* Partnered directly with researchers and scientists to translate ambiguous scientific workflows (e.g., DFT, GPAW materials science) into a concrete, functional product surface, making informed architectural decisions.
- *Data & Search Engineering:* Developed a hierarchical PDF reader and implemented advanced semantic search, significantly improving knowledge access by integrating diverse embedding models and generating candidate research questions.
- *System Optimization:* Optimized complex materials datasets stored on Figshare, achieving a substantial reduction in storage and transfer costs (from 2GB to 300Mb).
- *Agentic Pipelines:* Built an "Experiment agent" and validation ecosystem capable of ingesting historical research papers and attempting to synthesize final, high-quality research documents. Deployed multiple models for embeddings and qa on *GCP*
]
\
#resume-entry(
title: "Affogato (formerly Rendernet)",
location: "Bengaluru, India",
date: "12/2024 - 12/2025",
description: "Senior Software Engineer | Python, Golang, Postgres, Redis, Distributed Tracing",
)
\
#resume-item[
_Designing large-scale systems for media generation and delivery_
*System Architecture & Design Ownership*
- *Workflow DSL & Collaboration:* Designed an orthogonal, scalable architecture for a collaborative video editor, cleanly separating core concerns (timelines, assets, rendering). Developed a Workflow DSL backed by graph execution, enabling faster integration of complex media pipelines.
- *Concurrency & Indexing:* Architected fractional indexing primitives to support non-linear, collaborative editing in real-time, effectively eliminating global reindexing overhead and write contention issues.
- *Event-Driven Backbone:* Defined and owned asynchronous, event-driven patterns (queues, notifications) to decouple long-running media processing from user interactions, improving reliability and scalability.
*Platform Engineering & Tooling*
- *Developer Platform Ownership:* Standardized the entire internal developer experience (DevEx) by owning CI/CD standards, QA gates, and implementing automated PRD → ADR workflows using AI agents (Claude-Code / Opencode), significantly increasing developer velocity.
- *Authorization & Gateway Services:* Designed a team-based authorization system and the centralized reverse-proxy gateway with rate control and degraded-mode handling for external APIs, enhancing platform security and resilience.
- *Internal DSLs:* Built internal DSLs for cross-embedding, full-text search, and metadata enrichment to standardize retrieval and evaluation across products.
*Image & Video Optimization*
- *Image Pipeline:* Architected a high-throughput dynamic image delivery pipeline (Golang + libvips) with on-the-fly and background compression and dynamic worker scaling, reducing image transfer costs from ~4GB to ~10MB.
- *Upstream Contributions:* Drove upstream `libvips` and `govips` improvements by shifting to reader-based APIs, enabling streaming processing and automated C-bindings. #link("https://github.com/davidbyttow/govips/issues/476")[Upstream Issue]
- *Video Pipeline:* Designing a scalable ffmpeg farm for video compression, delivery, and editing workflows (with AWS MediaConvert as a later option).
*AI & Intelligence Capabilities*
- Built sophisticated agentic AI topologies, including an optimized LLM Router (shorter prompts + vector search) and internal tooling for rapid prompt experimentation, scaffolding entire product features.
- *Visual Intelligence:* Built a video analysis platform to extract screenplay structure, colors, and product metadata — foundation for knowledgebase refinement and campaign evaluation.
- Developed a robust critique system that combines LLM-based scoring, embedding scoring, and readability/emotional metrics for quality assessment.
*Development Velocity*
- Bridged the gap between product and dev teams for faster iteration on prompt changes; introduced convention changes to align frontend-backend deployment readiness, reducing post-release refactors.
]
\
#resume-entry(
title: "TheBackendCompany",
location: "Bengaluru, India",
date: "12/2023 - 12/2024",
description: "Freelance & Consulting",
)
\
#resume-item[
- Consulted multiple startups on early-stage architecture, infrastructure cost modeling, and strategic cloud adoption.
- Developed `go-batteries`, a reusable platform toolkit written in Go for scaffolding backend services and standardizing internal infrastructure.
- Built and deployed an LLM-backed API service using Ollama and LangChain for real-time text categorization and sentiment analysis experiments.
- Engineered a web-based CSV processing system leveraging SQLite WASM for fast, client-side analysis and transformation of large datasets.
]
\
#resume-entry(
title: "Sequoia Group",
location: "Bengaluru, India",
date: "08/2020 - 07/2023",
description: "Senior Software Engineer | Golang, Redis, AWS, DynamoDB, Python, gdb, valgrind, perf.",
)
\
#resume-item[
- *Feature Delivery:* Led development of multiple new features and addressed SSO login issues, centralizing CSV management across use cases for clients like Snap and Mongo, and handling file encoding issues.
- *Developer Platform Enhancement:* Built and maintained the internal developer platform — internal SDK, docs generation, database migrations, PII log filters, and a merge-conflict predictor — streamlining engineer workflows.
- *System Reliability & Observability:* Led migration of unencrypted files and integrated Datadog APM, systematically addressing slow queries and establishing centralized application and system metrics across multiple teams.
- *Security & Access Control:* Implemented and maintained the OpenID-based OAuth2 layer for secure employment history sharing, and developed robust API rate-limiting and validation mechanisms in collaboration with the Security Team. Resolved Redis connection issues to enable token caching and rate limiting.
- *Performance Engineering:* Improved core services through `pprof` benchmarking and Flamegraphs to identify and resolve deep-seated bottlenecks.
- *Core Service Ownership:* Developed and maintained critical services, including a notification system (AWS SQS/SES/SNS) and standardized configuration management via a Golang SDK for tracing and file encryption.
- *Other Notable Work:* Built a DB resolver library for Gorm V1, migrated code from Lambda to Airflow, and lectured on Go concurrency and Node.js Promises while assisting integration efforts (including Swagger doc generation).
]
\
#resume-entry(
title: "Gojek",
location: "Bengaluru, India",
date: "2016 - 2020",
description: "Product Engineer | Golang, Rails, Postgres, Redis, Lua, Kong",
)
\
#resume-item[
- *API Security & Gateway Migration:* Led the migration of the customer gateway from HAproxy to Kong (Nginx), significantly improving API security, optimizing load, and reducing AWS costs. This required deep learning of Lua for custom authentication plugins and advanced network analysis (`tcpdumps`).
- *System Decomposition:* Played a key role in the complex decomposition of the customer service monolith, involving implementing OAuth-based authentication, rebalancing providers, and managing the migration from Rails to Golang.
- *Global Search Enhancement:* Improved global search accuracy for e-commerce services by implementing SymSpell, enhancing fuzzy search capabilities and cross-selling logic based on user feedback.
- *API Whitelisting & Fraud Prevention:* Improved API whitelisting and fraud prevention, giving the security team greater visibility and control over public-facing APIs.
- *Disaster Recovery:* Contributed to the resolution of Kafka failures during a GCP live migration by adjusting critical system parameters (ZOO_TICK_TIME, API timeouts) to ensure stream leader stability, #link("https://issues.apache.org/jira/browse/KAFKA-4084")[kafka issue].
- *Internationalization:* Contributed to the i18n phase by adding location headers and localizing error messages in collaboration with translators.
- *E-commerce:* Played a key role in Gojek's e-commerce project, managing inventory synchronization and coordinating driver bookings across multiple services.
]
\
#resume-entry(
title: "Leftshift",
location: "Pune, India",
date: "08/2016 - 09/2016",
description: "Software Engineer | Nodejs, Javascript, React, Electron, Rails",
)
\
#resume-item[
- Worked on collecting and analyzing Gojek app reviews to provide actionable data insights for the data team. _(Project acquired by Gojek)._
]
#resume-entry(
title: "Kreeti Technologies",
location: "Kolkata, India",
date: "2014 - 2016",
description: "Software Engineer | Rails, Nodejs, React, Elixir",
)
\
#resume-item[
- Full-stack setup of Memento, a social network app — database design, MERN stack, Heroku deployment.
- Implemented login and account creation for CPG and Retailer flows in Rails (auth/authz) on smarketplace, with S3 + ImageMagick for photo upload and formatting, working alongside UI/UX engineers. #link("https://web.archive.org/web/20171030101912/https://www.smarketplace.com/")[preview]
- Built a commenting feature on the promotions page for multi-party communication, plus weekly/daily promotion reminders via ActiveJob and Sidekiq.
- Rebuilt the internal payslip generation with a dashboard in Elixir.
- Built a matrimony app with multi-tenancy (`act_as_tenant`), Devise-based auth, search, and an Angular frontend. #link("https://web.archive.org/web/20171030101912/https://merayog.com/")[preview]
]
= Education
#resume-entry(
title: "West Bengal University of Technology",
location: "Kolkata, India",
date: "2010 - 2014",
description: "B.Tech in Electronics & Communication Engineering",
)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment