Production systems engineered for scale, complexity, and reliability.
Our solutions showcase our technical approach: we identify computationally hard problems, architect robust systems, and optimize for production performance. These aren't prototypes—they're production systems serving thousands of users daily.
Slotly
Constraint optimization for academic scheduling
The Technical Problem
Academic timetabling is an NP-hard constraint satisfaction problem. You're simultaneously optimizing across multiple dimensions: faculty availability, room capacity, course prerequisites, student enrollments, lab equipment requirements, and administrative constraints. The solution space grows exponentially with institution size.
Most institutions solve this manually using spreadsheets—spending 2-4 weeks per semester on scheduling, only to discover conflicts after publication. Manual approaches fail because humans can't efficiently explore solution spaces with millions of potential combinations.
Our Solution
Slotly implements advanced constraint satisfaction algorithms with AI-powered optimization. The system:
- Models scheduling as a multi-objective constraint optimization problem
- Uses intelligent backtracking with conflict-directed search
- Provides real-time conflict detection with explanation
- Generates conflict-free schedules in minutes across multiple semesters
- Supports multi-tenant SaaS architecture with institutional isolation
Technical Stack
Who it's for
Autonomous colleges and universities handling complex scheduling requirements where manual methods fail at scale. Institutions needing robust, production-grade systems for mission-critical academic operations.
Jiffy
LLM orchestration for institutional knowledge retrieval
The Technical Problem
Educational institutions accumulate vast amounts of domain-specific knowledge—admission policies, course catalogs, fee structures, placement data—but lack systems to make that knowledge instantly accessible. Standard chatbots fail because they lack domain context. Generic LLMs hallucinate answers. Rule-based systems can't handle natural language variations across English, Hindi, and regional languages.
During admission cycles, institutions face query volume spikes (500+ inquiries/day) that overwhelm small teams. Delayed responses directly correlate with lost enrollments.
Our Solution
Jiffy implements retrieval-augmented generation (RAG) with fine-tuned language models:
- Custom knowledge base ingestion pipeline for institutional data
- Semantic search over vectorized document embeddings
- Context-aware response generation with source attribution
- Multilingual support (English, Hindi, 8+ regional languages) via translation layers
- Conversation state management for multi-turn interactions
- Confidence scoring with human escalation for low-confidence queries
Technical Stack
Who it's for
Indian universities and colleges with high admission query volume that need scalable, accurate, multilingual knowledge retrieval without headcount scaling. Institutions where response time directly impacts enrollment outcomes.
These systems demonstrate our engineering approach.
We tackle problems with inherent computational complexity, architect for production scale, and optimize for operational reliability. If you're facing similar challenges—problems where naive solutions fail, scale matters, and system reliability is non-negotiable—we should talk.
Discuss your technical challenges