All Gayatri Education Products & Platforms
A unified suite of institutional adaptive learning platforms, air-gapped financial engines, developer infrastructure, and operations automation — all verified with passing test suites and open code.
Gayatri AI Platform
Course-independent Socratic tutoring platform with 16-step pedagogical orchestration. Runs entirely on CPU with <1GB RAM via local SLM (llama-cpp-python) with zero cloud dependency. Scoped RAG ensures zero cross-course data contamination and absolute student privacy.
- 4 Dedicated Portals: Student Learning Desk, Teacher Copilot & Intervention Queue, Parent Oversight, and Multi-Tenant Institution Governance.
- 16-Step Pedagogical Turn Lifecycle: Enforces graduated Socratic questioning, misconception diagnosis, and zero answer leakage.
- Scoped Air-Gapped RAG: Institutions upload PDF, DOCX, or JSON; strictly isolated per-course retrieval with zero data leakage.
Gayatri Chemistry Tutor
NCERT-aligned adaptive AI tutor specialized for Class 11-12 NCERT, JEE, and NEET Chemistry (Thermodynamics & Inorganic Chemistry). Pairs a fine-tuned Qwen 2.5-1.5B model with deterministic formula and reaction evaluators.
- 17-Stage Fine-Tuning Curriculum: Trained on Colab T4 with Qwen2.5-1.5B-Instruct using LoRA/PEFT across thermodynamics and inorganic chemistry.
- Deterministic Answer Evaluators: Custom symbolic checks for numerical tolerances, stoichiometry, formula normalization, and MCQ without naive string matching.
- Root Misconception Diagnosis: Actively diagnoses conceptual flaws (e.g. standard enthalpy of formation, logic inversion) and directs remediation.
Gayatri AI BFSI Workbench
Institutional-grade financial operations workbench designed for Banks, NBFCs, and Insurers. Executes loan underwriting verification, health insurance claims adjudication, and RBI grievance triage behind strict air-gapped network guards.
- Strict Local Air-Gap Socket Guard: Intercepts operating system sockets to guarantee zero outbound transmission of sensitive customer financial records.
- Automated Loan Underwriting: Ingests scanned loan agreements, salary slips, and bank statements via RapidOCR, with arithmetic and reducing EMI validation.
- Health Insurance Claim Adjudication: Automated clause-level policy reconciliation against hospital bills, discharge summaries, and tariff schedules.
Gayatri Intern Workbench
Collaborative cloud-ready engineering IDE where interns develop in isolated Git worktrees, resolve conflicts visually, and collaborate with context-aware AI assistants to ship production PRs.
- Isolated Git Worktrees: Interns code in isolated file workspaces without branch collisions; features in-app merge conflict resolution.
- Monaco Side-by-Side Diff Editor: Visual review modal comparing original code with streaming AI suggestions before one-click patch application.
- In-Workspace Subprocess Test Runner: Collapsible terminal executing pytest and unittest directly in the worktree with 1-click 'Fix with AI' error piping.
Gayatri Agent SDK
Modular Python package providing standardized contracts, local execution harness, CLI runner, and synthetic test suites for building modular AI agents across Gayatri's 47-agent mesh.
- Standardized Agent Contract: Enforces BaseAgent, AgentContext, and AgentResult conventions across all cognitive clusters.
- CLI Offline Runner: `gayatri-agent` CLI tool for testing agent reasoning offline with synthetic test payloads and confidence scoring.
- Powers 47-Agent Mesh: Reusable substrate for Researcher, Document Analyst, Code Reviewer, Policy Evaluator, and Orchestrator agents.
Gayatri Local LLM API
Drop-in, 100% OpenAI-compatible local API server using local GGUF models via llama-cpp-python and FastAPI. Features dynamic RAM allocation, LRU model eviction, and native tool calling.
- 100% OpenAI Specification: Implements /v1/chat/completions, /v1/models, and /v1/embeddings. Works seamlessly with official OpenAI SDKs.
- Dynamic VRAM/RAM Memory Management: Automatically calculates memory requirements from GGUF metadata with LRU eviction and idle unloading.
- Native Tool Calling: Translates OpenAI function schemas into GGUF grammar constraints or XML routing (<tool_call>) for models like Qwen 2.5.
GayatriBot HR Automation
Enterprise-grade automated HR and recruitment system connecting Google Forms, local Ollama resume parsing, rotating SMTP email pools, automated PDF offer letters, and administrative dashboards.
- Continuous Form Polling: Monitors incoming applicant submissions every 5 minutes via Google Sheets API.
- Local AI Resume Parsing: Uses local Ollama models to parse, extract skills, and match candidate CVs against internship prerequisites.
- Rotating Multi-Account SMTP Pool: Rotates outbound dispatch across sender accounts with warmup algorithms to prevent rate-limiting and spam flags.
Gayatri Referral & Growth System
Dedicated Flask application tracking student and intern referrals with 30-day attribution cookies, roster-gated authentication, dual-stage admin payment verification, and real-time earnings dashboards.
- Roster-Gated Authentication: Candidate signups are cross-referenced with approved cohort rosters before granting referral privileges.
- 30-Day Cookie Attribution: Tracks referral links with last-click attribution resolution and HTTP-only cookie persistence.
- Dual-Gate Accounting Verification: Separates applicant approval from payment screenshot verification for rigorous compliance.
Gayatri STEM Knowledge Packs
Turnkey institutional curricula engineered for scoped RAG ingestion: MATH201 (Engineering Mathematics), EC202 (Digital Electronics), and CHEM101 (Senior Secondary Chemistry) with pre-indexed concept graphs.
- MATH201 (Engineering Math): ODEs, Laplace Transforms, Fourier Series, and Linear Algebra with 30+ concept cards.
- EC202 (Digital Electronics): Boolean algebra, logic gates, flip-flops, with detection & remediation of logic inversion traps.
- CHEM101 (Chemistry): Thermodynamics and molecular structure mapped to NCERT and JEE competitive standards.
Catalyser Dedicated IIT-JEE Tutor Model
Authoritative technical and commercial specification (DBERT x Catalyser Strategic Partnership, Indore) defining dedicated on-premise model weights, milestone delivery, and adaptive learning systems for IIT-JEE Advanced.
- Operational Hub in Indore, Madhya Pradesh (Catalyser Strategic Partnership: https://www.catalyser.in/).
- Dedicated adaptive learning models tailored for IIT-JEE Main, Advanced & Olympiad problem solving.
- Strict Two-Pillar Knowledge vs Behavior governance with symbolic math and physics evaluators.
Catalyser Dedicated IIT-JEE Tutor Model
Authoritative technical and commercial specification (DBERT x Catalyser Strategic Partnership, Indore) defining dedicated on-premise model weights, milestone delivery, and adaptive learning systems for IIT-JEE Advanced.
- Operational Hub in Indore, Madhya Pradesh (Catalyser Strategic Partnership: https://www.catalyser.in/).
- Dedicated adaptive learning models tailored for IIT-JEE Main, Advanced & Olympiad problem solving.
- Strict Two-Pillar Knowledge vs Behavior governance with symbolic math and physics evaluators.
The Two Pillars: Knowledge vs. Behavior
To eliminate hallucinations and prevent data leaks, our systems strictly divide cognitive work into two layers:
1. The Knowledge Layer (Deterministic & Scoped)
Curriculum syllabi, insurance policy clauses, and loan agreements are stored in scoped, version-pinned vector databases and relational tables. The AI is forbidden from answering without explicit citation.
2. The Behavior Layer (Neural & Pedagogical)
Fine-tuned small language models (Qwen 2.5) act exclusively as pedagogical guides and logical reasoners — choosing whether to provide an analogy, diagnostic question, or stepped hint without inventing facts.
Deploy Gayatri AI in your institution
Whether you are a university looking for on-premise adaptive tutoring or a financial institution needing air-gapped claim adjudication, our engineering team can deploy a dedicated pilot.