SHS
Resume

AI systems engineering / backend / open source

Sri Hari Haran
Sharma

I’m a builder who enjoys understanding how things work under the hood and turning ideas into working systems.

I work across AI, backend engineering, data, databases, JSON Schema, and open source, solving real-world problems and building things that matter.

I am interested in researching ideas and applying research through working software. I also use simulation software to make complex behavior visible, test assumptions and understand systems better.

Having no other choice is its own blessing.
GitHub LinkedIn Hashnode

Recognition

03

Hackathon Winner at Deccan AI Catalyst 2026

Selected from a solo AI-agent hackathon with more than 1,500 registrations and 500 working agents. I built Sizuka, an AI recruiter agent for finding, evaluating and reaching out to candidates.

View the result

Experience

Work experience

Freelance · Remote, UKOct 2026 - Present

Full-Stack Engineer

Shoowai

Develop and maintain full-stack marketplace features using React, TypeScript, Supabase and Cloudflare. Implement secure database access, authentication workflows, payment integrations and social-sharing metadata.

  • React
  • TypeScript
  • Supabase
  • Cloudflare
Internship · HybridJun - Sep 2026

Applied AI Engineer

The Intersections

Built and published ilx-datatable, a reusable Svelte data-table package. Owned the end-to-end architecture and agent-assisted workflow for VMPire.

  • Svelte
  • Architecture
  • AI workflows
Freelance · RemoteJan - Mar 2026

AI Agent Evaluator

Deccan AI Experts

Evaluated multi-step agent workflows, investigating tool-selection, execution and coordination failures and providing structured feedback to improve agent performance.

  • Agent evaluation
  • RLHF
  • Tool use

01

AI, ML and deep learning projects

Agent systemFeatured

Sizuka

An AI talent scouting agent that turns a job description into a live candidate discovery and outreach workflow.

Sizuka extracts hiring criteria, discovers developers through GitHub and Hacker News, analyses profiles, scores candidate fit with an LLM, generates personalised outreach and follows email replies. NestJS, BullMQ and Redis coordinate the workflow while a Python agent, PostgreSQL and a React dashboard handle intelligence, state and visibility.

  • LangGraph
  • NestJS
  • LLM evaluation
  • PostgreSQL
  • Redis
View repository
Applied AIMobile

Pramana

An AI-powered fact verification application designed for one-tap checking from the Android share sheet. It returns an explanation with traceable citations so the result can be inspected instead of merely trusted.

View repository
JSON SchemaResearch prototype

Ecosystem observability

A small evidence pipeline for measuring JSON Schema adoption, validator compliance, draft evolution and ecosystem breadth. It combines package, repository and Bowtie signals into historical data that can be compared over time.

View repository

02

Open source and JSON Schema

BlazeMoved unevaluated cycle tests into draft suitesPR 1065 BlazeTested cyclic references in dependency analysisPR 1056 Twenty CRMCorrected MCP response contracts and HTTP semanticsPR 18671 Twenty CRMFixed orphan navigation items for deleted viewsPR 18791 Twenty CRMSynchronized record fields arriving through SSEPR 19069 Twenty CRMFixed MCP streamable HTTP method handlingPR 19496 Twenty CRMFixed REST filter conjunction detectionPR 20133 JSON SchemaAdded weekly ecosystem statistics collectionPR 26 JSON SchemaImproved Test Suite documentationPR 867

03

Writing

AI engineering / documentationHow I Stopped My AI Coding Agents from Turning My Codebase into SoupRead article JSON Schema / AI agentsWhy JSON Schema Is the Missing Contract in AI Agent CommunicationRead article Software architectureOne Application, Three ArchitecturesRead article Backend foundationsFrom DNS to Localhost: A Practical Guide to Backend FoundationsRead article

04

Books that shape how I think

Technical

  • Mathematics for Machine LearningMarc Peter Deisenroth, A. Aldo Faisal and Cheng Soon Ong
  • Designing Data-Intensive ApplicationsMartin Kleppmann
  • Database Management SystemsRaghu Ramakrishnan and Johannes Gehrke

Beyond engineering

  • Think Like a Maths GeniusArthur Benjamin and Michael Shermer
  • The Theory of EverythingStephen Hawking
  • The AlchemistPaulo Coelho
  • I'm OK, You're OKThomas A. Harris

Current direction

Research an idea. Build the smallest useful version. Simulate the hard parts. Measure what changes.