H.HARSH SINGH

PROJECT PORTFOLIO · SELECTED WORK

Systems built to
solve real problems.

A curated view of research infrastructure, agentic engineering tools, quantitative systems, creative technology, and applied products. Each project is described by its actual scope—not by inflated claims.

22PROJECTS
17PUBLIC REPOSITORIES
5PRIVATE PROJECTS
1UPSTREAM REFERENCE

Filter by discipline. Public projects link to source repositories. Private projects are summary-only and have no source links.

01 / AI RESEARCH, REASONING & MACHINE LEARNING07 PROJECTS
MULTI-AGENT RESEARCH INFRASTRUCTUREPUBLIC

Argumentum Research Orchestra

Self-hosted research infrastructure coordinating literature discovery, evidence mapping, methodological critique, manuscript workflows, and human-governed experiments.

10 SPECIALIST ROLES · 13 RESEARCH STAGES · SELF-HOSTED
View repository ↗
RESEARCH WORKFLOW TOOLKITPUBLIC

ReasoningAI Studio

Evidence-first research toolkit with portable workflows, research prompts, templates, and local Python utilities for literature work, evaluation, and reproducibility.

64 SKILLS · RESEARCH UTILITIES · REPRODUCIBILITY
View repository ↗
INTERDISCIPLINARY RESEARCH PLATFORMPRIVATE PROJECT

EPISTEME Research Studio

Research workflow system bringing together evidence-grounded methods, specialist roles, evaluation procedures, research artifacts, and integrity checks.

88 RESEARCH SKILLS · 56 SPECIALIST ROLES · V0.6.0
Private development · Source repository not linked
SEARCH & REASONING EVALUATIONPRIVATE PROJECT

ReasonTree

Local-first, model-agnostic Tree-of-Thought search and evaluation framework with bounded exploration, multiple search policies, task verifiers, and experiment analysis.

BFS · DFS · BEAM SEARCH · MCTS
Private research software · Source repository not linked
CONTROLLED LLM EXPERIMENTSPRIVATE PROJECT

CoTCompare

Research workbench for paired comparisons of direct-answer and stepwise-prompting conditions, with structured outputs and an answer-key evaluation harness.

RESEARCH PROTOTYPE · 6 STARTER EXAMPLES
Experimental prototype · No general reasoning-improvement claim
PROMPT & CONTEXT EFFICIENCYPRIVATE PROJECT

Token Saver

Selective prompt-refinement and context-efficiency toolkit that evaluates the total cost of reaching an accepted result, including refinement overhead, retries, latency, and output.

120 EVALUATION TASKS · 4 MODEL FAMILIES
Private development · Evaluation claims remain tied to documented runs
INTERACTIVE MACHINE LEARNINGPUBLIC

Neural Networks Playground

Visual-first machine-learning lab for training small neural networks and inspecting decision boundaries, activations, weights, gradients, and optimiser behavior.

BROWSER MLP · NUMPY · EXPERIMENT COMPARISONS
View repository ↗
02 / AGENTIC ENGINEERING & DEVELOPER SYSTEMS04 PROJECTS
CROSS-AGENT ENGINEERING TOOLKITPUBLIC

AI Agent Engineering Toolkit

Curated, portable agent skills, specialist role cards, and workflow templates for software engineering, architecture, debugging, testing, research, and technical delivery.

50 SKILLS · 10 ROLES · 5 TEMPLATES
View repository ↗
CODEX-FIRST DELIVERY SYSTEMPUBLIC

LaunchCrew

Spec-first engineering toolkit that turns product ideas into scoped requirements, delegated implementation, staged verification, independent review, and release handoffs.

23 SKILLS · 10 CODEX SPECIALISTS
View repository ↗
CLAUDE CODE PLUGINPUBLIC

Claude Garage

Claude Code plugin providing a startup-style specialist crew and resumable workflows for discovery, planning, parallel implementation, testing, security review, and shipping.

13 SPECIALIST AGENTS · 20 COMMANDS
View repository ↗
SOFTWARE ARCHITECTURE & REFACTORINGPUBLIC

Implicit State Machines

Portable workflows for detecting hidden application lifecycles and safely modelling boolean-heavy state as explicit, testable transitions.

2 FOCUSED SKILLS · STATE MODELLING
View repository ↗
03 / QUANTITATIVE FINANCE & MARKET SYSTEMS03 PROJECTS
TRADING RESEARCH & RISKPUBLIC

MarketPilot AI

Cross-model trading research toolkit for market scans, scenario replay, walk-forward evaluation, position sizing, deterministic risk checks, and a local paper ledger.

PAPER-ONLY · RISK CHECKS · RESEARCH ENGINE
View repository ↗
MULTI-SPECIALIST MARKET RESEARCHPRIVATE PROJECT

MarketCouncil

Trading research council coordinating specialist analysis, bull/bear thesis challenge, evidence review, deterministic risk utilities, and human-owned decisions.

15 AGENT ROLES · 74 SKILLS · 13 WORKFLOWS
Private development · Research tool, not an automated trading service
QUANTITATIVE & MARKET MICROSTRUCTUREPUBLIC

JaneStreet Agentic Skills

Specialist playbooks covering order-book reconstruction, market making, execution analysis, low-latency systems, risk controls, and quantitative research validation.

MARKET MICROSTRUCTURE · EXECUTION · RISK
View repository ↗
04 / CREATIVE COMPUTING, DESIGN & SIMULATION04 PROJECTS
CREATIVE PRODUCTION TOOLINGPUBLIC

AI Creative Production Toolkit

Portable agent skill library and local Python toolkit for creative workflows across image, video, audio, 3D, game art, asset operations, and production QA.

219 AGENT SKILLS · PYTHON TOOLKIT
View repository ↗
INTERFACE QUALITY & DESIGN SYSTEMSPUBLIC

Pixelcraft

Specialized skill library for identifying generic or inconsistent interface patterns and improving hierarchy, typography, layout, accessibility, content, and design-system coherence.

51 DESIGN-QUALITY SKILLS · UI REVIEW
View repository ↗
MECHANICAL SIMULATION & VISUALISATIONPUBLIC

KINETIQ Machine Simulation Skills

Portable workflows for interactive 3D machine explainers, mechanical reconstruction, kinematics, physics validation, consistent readouts, and offline delivery.

17 SKILLS · PHYSICS VALIDATION
View repository ↗
ANIMATED SHORT PRODUCTIONPUBLIC · RELEASE CANDIDATE

Kinoquence

Local-first, model-agnostic workflow for turning AI-assisted story ideas into structured short-video projects with storyboards, assets, captions, credits, and quality checks.

PYTHON WORKFLOW TESTED · REMOTION RENDER NOT YET VERIFIED
View repository ↗
05 / APPLIED PRODUCTS, WORKFLOWS & SECURITY04 PROJECTS
STARTUP EVALUATION FRAMEWORKPUBLIC

The Sharks

Multi-agent startup evaluation panel that challenges customer demand, revenue assumptions, unit economics, competitive position, and founder-market fit.

5 SPECIALIST AGENTS · 6 SKILLS
View repository ↗
PROFESSIONAL WORKFLOW ECOSYSTEMPUBLIC

LinkedIn Skills 360

Model-agnostic skill ecosystem for professional identity, job search, networking, content, sales, recruiting, hiring, and career workflows.

141 SKILLS · 18 SPECIALIST AGENTS · 32 TEMPLATES
View repository ↗
AUTHORIZED API INTEGRATIONPUBLIC

SheerFlow — SheerID Workflow Studio

JavaScript workbench for authorized SheerID REST API integrations, verification lifecycles, OAuth handling, document uploads, job orchestration, and auditable operations.

NODE.JS · OAUTH 2.0 · API WORKFLOWS
View repository ↗
DEVELOPER SECURITY TOOLINGPUBLIC

HoneyPot Guard

Local-first pre-commit scanner for exposed credentials, suspicious outbound destinations, and risky code patterns, with redacted findings and configurable blocking thresholds.

NODE.JS · ZERO RUNTIME DEPENDENCIES
View repository ↗
ECOSYSTEM REFERENCE · NOT ORIGINAL AUTHORSHIP

MAPIE — Model Agnostic Prediction Interval Estimator

A repository copy of the upstream scikit-learn-contrib project for conformal prediction, uncertainty quantification, and risk control. Included as a technical ecosystem reference, not presented as an independently developed project.

View upstream-based repository ↗

Private projects are presented as high-level summaries only; no private repository links or source files are exposed. Status labels distinguish public projects, private development, and release candidates. Descriptions reflect documented scope and do not imply production deployment, investment performance, or independently validated scientific findings.

PUBLIC SOURCE CODE & PROJECT HISTORYBrowse public GitHub repositories ↗