Ai Engineer

About Drishti IAS Drishti IAS is one of India’s most trusted institutions for UPSC Civil Services preparation, serving lakhs of aspirants through classroom programmes, a bilingual (Hindi and English) digital platform, daily current-affairs content, test series and video learning. We are bringing intelligent, reliable and multilingual learning experiences to our students and making our editorial, operations and support workflows dramatically more efficient through AI. About the Role We are looking for a hands-on AI Engineer with 1–3 years of experience who can take LLM-based systems from idea to production. You will split your time roughly equally between two tracks: student-facing AI products such as doubt-resolution assistants, answer-writing evaluation, personalised study aids and voice/vision-enabled learning; and internal automation that accelerates content creation, data pipelines and business operations. You will own the full stack of an AI feature: data, retrieval, models, serving, evaluation and monitoring. Key Responsibilities LLM & RAG systems: Design, build and maintain retrieval-augmented pipelines over Drishti’s large content corpus, including notes, current affairs, PYQs and test series. Own chunking, embedding, hybrid search, re-ranking and citation quality, and set up offline/online evaluation and production monitoring for accuracy, hallucination and latency. Agentic workflows & automation: Build tool-using agents with LangChain/LangGraph or similar SDKs, expose capabilities via MCP, and automate editorial, marketing and support workflows on platforms such as n8n, Dify or Zapier. Inference & deployment: Deploy and serve open-weight models on GPU infrastructure using vLLM, Transformers or llama.cpp. Handle quantisation, batching, KV-cache and throughput, latency and cost trade-offs. Fine-tuning: Create high-quality datasets from in-house material and fine-tune LLMs using SFT, LoRA, QLoRA and preference tuning for domain, tone and Hindi-English tasks. Evaluate rigorously against baselines before release. Multimodal integration: Integrate speech tools for STT/TTS, including Whisper and ElevenLabs, for voice-based learning and dictation. Integrate image-generation and editing models, including diffusion models and Nano Banana, for content and creative workflows. Data engineering: Build pipelines that ingest and process large volumes of text, audio and long documents. Design schemas and queries across PostgreSQL/MySQL and vector databases such as pgvector, Qdrant and Milvus. System design & reliability: Design scalable, observable services with clean APIs. Add logging, tracing, cost tracking and guardrails. Apply secure-by-default practices, including prompt-injection defences, secrets management and least-privilege access for tools and agents. Collaboration: Work closely with subject-matter experts, editors and product stakeholders, and document systems clearly. Required Skills & Experience Python: Strong, production-quality Python skills, including typing, async, packaging and testing. Comfortable with FastAPI or a similar framework. Automation platforms: Hands-on experience with n8n, Dify, Zapier or comparable low-code/workflow tools, including custom nodes and webhooks. RAG end-to-end: Experience building, evaluating and monitoring RAG systems using RAGAS, custom evaluations or LLM-as-a-judge approaches. Strong SQL skills with PostgreSQL/MySQL and practical experience with vector databases. Agent frameworks: Experience with LangChain, LangGraph or similar SDKs/ADKs, such as Google ADK, OpenAI Agents SDK or Pydantic AI. Knowledge of function/tool calling and Model Context Protocol (MCP). System design: Ability to design and reason about services, queues, caching, data flow and failure modes for AI applications. Inference engineering: Experience deploying local/open-weight models using vLLM, Transformers or llama.cpp. Knowledge of GPUs, VRAM budgeting, quantisation formats such as GGUF, AWQ and GPTQ, and CUDA basi