⚡ Foundation Models & RAG · DBERT ACADEMIC SPECIFICATION 2026
Generative AI & LLM Systems Engineering
₹6,50,000 – ₹18,00,000 LPA Verified Average Remuneration · Performance Stipend: ₹5,000 fixed + up to ₹18,000 performance incentive
Learn parameter-efficient fine-tuning (LoRA/QLoRA), dense vector search, Reciprocal Rank Fusion, and production LLM orchestration on real venture codebases.
PyTorchTransformersLoRA / QLoRALangChainFastAPIRAG
Program Structure Overview
Structure 1: Fellowship
2 Months (8 Weeks)
Fee: ₹1,999 (₹499 deposit + ₹1,500 balance) · ₹5,000 stipend release
Structure 2: Paid Course
3 Months (12 Weeks)
Comprehensive foundational to enterprise mastery · Zero-cost EMI
2-Month Fellowship Sprint Milestones
Mandatory 10.0 hours/week logged on the platform attendance daemon with upstream PR contributions.
Weeks 1–2Tokenization & Vector Embedding Pipelines
- Build dense vector search with cosine similarity
- Implement hybrid BM25 + dense search
- PR on Alkame data fetcher
Weeks 3–6Fine-Tuning & Evaluation Sprints
- Fine-tune Mistral/Llama with QLoRA on domain datasets
- Build RAG evaluation benchmark with Ragas
- Deploy FastAPI streaming inference
Weeks 7–8Latency Benchmarking & Credential Release
- Optimize KV-cache memory and GPU batching
- Submit Turnitin-verified research dossier
- Unlock ₹5,000 stipend release
3-Month Comprehensive Course Syllabus
12 structured weekly modules, hands-on lab exercises, and 2 enterprise production capstones.
Weeks 1–4Transformers & Embedding Foundations
- Attention mechanisms
- Sentence transformers
- Vector database schemas
Weeks 5–8Advanced RAG & Prompt Chaining
- Reciprocal Rank Fusion
- Context window pruning
- Hallucination guardrails
Weeks 9–12PEFT Fine-Tuning & Enterprise Production
- LoRA/QLoRA pipelines
- Model quantization (AWQ, GGUF)
- Production deployment on AWS
Official Upstream Partner Repository
Fellows contribute production code directly to Alkame AI Market Intelligence. Personal forks without upstream PRs are rejected by the CI grading daemon.
Scope: Financial sentiment NLP with FinBERT, time-series embeddings, and streaming analysis.
Admissions & Curriculum FAQs
Do I need access to an expensive GPU?
No. DBERT provides cloud sandboxes and Google Colab Enterprise GPU runtimes for all fine-tuning and inference labs.
Official Admissions Ingestion
Enroll in Generative AI & LLM Systems Engineering
Submit details below to receive the complete syllabus breakdown, cohort standup schedule, and admission counseling.