Kerdos Infrasoft Private Limited is an AI Safety & Alignment Infrastructure company building the foundational systems needed to train superintelligence safely and responsibly. Founded in Bengaluru in December 2023, we are pioneering infrastructure, alignment systems, and interpretability tools that enable frontier AI research while prioritizing safety, transparency, and responsible development practices.

"Building infrastructure to train superintelligence safely."

Build and scale distributed training systems for frontier AI models.

Capabilities:

• Distributed Compute Orchestration
• Model Scaling Frameworks
• Efficient Training Pipelines
• Fault-Tolerant Architecture
• Multi-Node Synchronization
• Exascale Training Systems

Tech Stack: PyTorch Distributed · DeepSpeed · Megatron-LM · Ray · Kubernetes

Ensure AI systems align with human values and operate safely.

Capabilities:

• Constitutional AI Training
• Reward Modeling Systems
• Preference Learning Pipelines
• Value Alignment Frameworks
• RLHF at Scale
• Safety Evaluation Suites

Tech Stack: Anthropic Claude · OpenAI Alignment · Constitutional AI · RLHF · Debate

Understand and monitor AI model behavior at scale.

Capabilities:

• Mechanistic Interpretability
• Activation Analysis
• Circuit Discovery
• Feature Visualization
• Model Behavior Monitoring
• Anomaly Detection

Tech Stack: TransformerLens · Captum · InterpretML · SHAP · Neuron Viewer

Infrastructure and tools for cutting-edge AI research.

Capabilities:

• Research Collaboration Tools
• Governance Frameworks
• Safety Evaluation Systems
• Reproducibility Infrastructure
• Peer Review Platforms
• Open Science Practices

Tech Stack: Weights & Biases · MLflow · DVC · Papers with Code · ArXiv

Private Document Q&A & LLM Chat powered by RAG and LLaMA 3.1

Key Features:

• Private LLM deployment - zero data leaves your environment
• Document Q&A with RAG architecture
• FAISS vector search for semantic retrieval
• Works with PDF, DOCX, TXT, CSV, MD formats
• CPU-friendly - no GPU required

Benefits:

• Complete data privacy and security
• On-premise or private cloud deployment
• Customizable for enterprise workflows
• Fast document processing and retrieval

Try the Demo: huggingface.co/spaces/kerdosdotio/Custom-LLM-Chat

Official Python SDK for Kerdos AI services

Key Features:

• Easy Integration with existing workflows
• Type Hinting for better developer experience
• Async Support for high-performance applications
• Extensive Documentation and examples

Benefits:

• Programmatic access to Kerdos AI capabilities
• Build custom agentic workflows
• Rapid LLM prototyping
• Production-ready code

Solutions delivered:

• Red-teaming platforms
• Adversarial testing suites
• Safety evaluation frameworks
• Threat modeling tools

Solutions delivered:

• Constitutional AI frameworks
• Preference learning systems
• Reward modeling platforms
• Value alignment tools

Solutions delivered:

• Mechanistic interpretability tools
• Circuit discovery frameworks
• Activation analysis systems
• Feature visualization platforms

Solutions delivered:

• Governance frameworks
• Safety standard protocols
• Oversight systems
• Compliance monitoring tools

• Deep Learning: PyTorch Distributed, DeepSpeed, Megatron-LM, TensorFlow
• LLMs: LLaMA 3.1, Anthropic Claude, OpenAI GPT, Constitutional AI
• Vector Search: FAISS, Pinecone
• ML Ops: Weights & Biases, MLflow, DVC
• Interpretability: TransformerLens, Captum, InterpretML, SHAP

• Framework: Next.js 16.2.3 (App Router)
• Language: TypeScript 5
• React: 19.2.3
• Styling: Tailwind CSS v4 with custom OKLCH color system
• UI Components: shadcn/ui, Radix UI primitives
• Animations: Framer Motion 12.34.0
• 3D Graphics: Three.js 0.183.2

• Orchestration: Kubernetes, Ray
• Cloud: AWS, Azure, Google Cloud
• Storage: AWS S3, Upstash Redis, Upstash KV
• Authentication: NextAuth v5, SimpleWebAuthn (passkeys)
• Email: Mailtrap, Nodemailer

• Password Hashing: bcryptjs
• Sanitization: DOMPurify
• Rate Limiting: Upstash Redis
• Headers: HSTS, CSP, X-Frame-Options

• Analytics: Vercel Analytics
• Monitoring: Web Vitals tracking
• Optimization: Sharp (images), Critters (critical CSS)
• Caching: Static asset caching (1 year)

• Linting: ESLint 9 (flat config)
• Type Safety: TypeScript strict mode
• Dev Server: Turbopack
• Workers: Wrangler (Cloudflare Workers)

• Display Font: Montserrat (hero headings)
• Sans Font: Lexend (body text, UI)
• Accessible Font: Atkinson Hyperlegible (labels, captions)
• Monospace Font: JetBrains Mono (code blocks)

• Primary: oklch(55% 0.22 250) - deep corporate blue
• Accent: oklch(70% 0.18 230) - interactive blue
• Surface: oklch(20% 0.04 250) - dark card background
• Background: oklch(14% 0.02 250) - near-black
• Foreground: oklch(95% 0.01 250) - off-white

• Service worker for offline functionality
• Install banner for native-like experience
• Offline fallback page
• Manifest with app icons

• Skip-to-content links
• Semantic HTML structure
• Atkinson Hyperlegible font for readability
• Keyboard navigation support
• Screen reader optimization
• ARIA labels and live regions

• Structured data (JSON-LD)
• OpenGraph and Twitter cards
• Dynamic sitemap and robots.txt
• 4.3★ rating display (AmbitionBox)
• Job posting schema for careers
• Local business schema

• LCP (Largest Contentful Paint): < 2.5s
• FID (First Input Delay): < 100ms
• INP (Interaction to Next Paint): < 200ms
• CLS (Cumulative Layout Shift): < 0.1
• TTFB (Time to First Byte): < 600ms
• Lighthouse Score: Target > 90

• Bottom navigation (Home, Services, Quote, Contact, More)
• Safe area inset support (iPhone notch)
• Haptic feedback
• Pull-to-refresh controls
• Touch-optimized targets (44×44px minimum)
• Responsive breakpoints

We're actively hiring across AI safety, alignment research, platform engineering, and more. Join us in building the infrastructure for safe superintelligence. All positions are based in Bengaluru with hybrid/remote options available.

We're looking for passionate individuals who want to work on cutting-edge AI safety and alignment challenges. If you're excited about:

• Training superintelligence safely
• Building AI alignment infrastructure
• Mechanistic interpretability research
• AI governance and policy

Apply now: careers@kerdos.in or visit kerdos.in/careers

© 2024–2026 Kerdos Infrasoft Private Limited. All rights reserved.

This document is provided for informational purposes. Kerdos Infrasoft Private Limited is a duly incorporated company under the Companies Act, 2013.

CIN: U62099KA2023PTC182869  ·  DPIIT Startup India: DIPP170654  ·  UDYAM (MSME): UDYAM-KR-03-0393965