I build
systems
.

Building systems

that outlast the chaos

Who I Am

Subin Joshua Sunil

Technical leader and systems architect — I walk into chaos and ship infrastructure that outlasts it.

Director of Technology & Operations at ThinkFISH, where I lead a 20+ member team and own the entire operational infrastructure — platforms, data warehouse, automation pipelines, and five production websites serving 108+ B2B clients.

I've built three production platforms from scratch, solo. An enterprise analytics tool processing 480+ campaigns. A data warehouse handling 48K prospects at 99.9% uptime. A LinkedIn outreach operating system with seven phases of development complete.

My pattern: find the bottleneck, build the system, document the playbook, move to the next one.

Platforms built (solo)3
B2B clients served108+
Active campaigns480+
Team managed20+
Newsletter subscribers86K+
Production uptime99.9%
108+ Clients·480+ Campaigns·48K+ Prospects·86K+ Subscribers·99.9% Uptime·36K Lines TypeScript·3 Platforms·1st Place GTM Hackathon·108+ Clients·480+ Campaigns·48K+ Prospects·86K+ Subscribers·99.9% Uptime·36K Lines TypeScript·3 Platforms·1st Place GTM Hackathon·
Experience
2024 — Present

Director of Technology & Operations

ThinkFISH

Lead 20+ member team, own full operational infrastructure, reduced manual work 80%, scaled newsletter 0 → 86K subscribers. Primary architect across ARGUS, Data Warehouse, and all automation systems.

2023 — 2024

Systems Manager

ThinkFISH

Unified 5+ fragmented platforms, led HubSpot CRM implementation post-acquisition. Designed and shipped the operational infrastructure that now runs the company. Promoted to Director within 8 months.

2022 — 2023

Operations & Automation Lead

Manhattan Global Partners

Automated workflows across CRM, research databases, and reporting pipelines — 70% reduction in manual effort. Built operational dashboards giving leadership real-time visibility.

2021 — 2022

Operations Specialist

RealtySimplified LLC

Built lead management workflows with automated routing, follow-ups, qualification, and status tracking. Implemented real-time alerting for pipeline health monitoring.

2018 — 2021

Technical Support Lead

HP Inc., Bangalore

Led customer support operations team. Built macro toolkit replacing manual copy-paste workflows. Created knowledge base system for recurring issues. Automated weekly reporting with Python.

What I Work With

Technical
Skills.

Languages & Frameworks

TypeScriptJavaScriptPythonNode.jsNext.js 16React 19NestJSExpress

Databases & Infrastructure

PostgreSQLPrismaDrizzle ORMRedisBullMQDockerNginxHetznerBare Metal

AI & Automation

Claude APIOpenAI / GPTPerplexity SonarLangChainn8nZapierMake.comInngest

GTM & Operations

ClayInstantlySmartleadHubSpotGoHighLevelCalendlyUnipileValley

Architecture & Practices

Multi-tenant SaaSREST APIsWebhooksReal-time (SSE)Background JobsRBACPrometheusCI/CD
Beyond the Stack
01

Designing Websites

Dark aesthetics, type-driven layouts, scroll-driven interactions — I care about how things feel, not just how they work.

02

Building Scrapers

Structured data extraction at scale. Headless browsers, proxy rotation, anti-detection patterns, pipeline orchestration.

03

Cryptography

Encryption schemes, hashing algorithms, zero-knowledge proofs, protocol design — the math that keeps systems honest.

Achievements
3 Platforms
Built solo in production
ARGUS · LinkDone · Data Warehouse
8 Months
Systems Manager → Director
ThinkFISH · Fastest promotion in company history
86K+
Newsletter subscribers
Scaled from zero · Highest-converting lead channel
What's Next

ML From First Principles

I can ship production AI apps — but calling claude.messages.create() isn't the same as understanding what's underneath. Currently grinding through the math so I can build models, not just consume them.

01

Linear Algebra & Matrix Theory

Eigendecomposition, SVD, matrix calculus, vector spaces — the language every model actually speaks. Not just numpy.linalg, but understanding why PCA works and what attention heads are really computing.

02

Probability, Statistics & Information Theory

Bayesian inference, maximum likelihood, KL divergence, entropy, conditional distributions. The math behind why cross-entropy loss works, not just that it does.

03

Optimization & Calculus

Gradient descent variants, backpropagation derivations, convexity, loss landscapes, learning rate dynamics. Building intuition for why models converge — or don't.

04

From Theory to Architecture

Implementing transformers, attention mechanisms, and training loops from scratch in PyTorch. No wrappers, no HuggingFace — raw tensor operations to bridge the gap between math and working models.

05

Applied ML Engineering

Fine-tuning, RAG architectures, evaluation harnesses, RLHF pipelines, and model deployment. The endgame: not just consuming AI APIs, but building the systems behind them.