Generative AI Internship | Spypro Hack You

Program Overview

Real Models. Real Pipelines.
Real Products.

This isn't a theoretical AI survey. From day one you'll be engineering prompts on frontier LLMs, fine-tuning open-source models on domain data, building diffusion-based image workflows, and integrating Gen AI into production applications alongside experienced AI engineers.

We built this program around what employers actually need: solid prompt engineering discipline, fine-tuning intuition, safety and bias awareness, and the ability to ship reliable Gen AI features into real products.

3-4 months

Remote & hybrid

Certificate

Part-time ok

Live LLM projects

finetune_llm.py spypro-genai

            from transformers import AutoModelForCausalLM, TrainingArguments
            from peft import LoraConfig, get_peft_model
            from trl import SFTTrainer
            

            # Load base model + LoRA adapter
            model = AutoModelForCausalLM.from_pretrained(
              "mistralai/Mistral-7B-v0.1", load_in_4bit=True
            )
            lora_cfg = LoraConfig(r=16, lora_alpha=32, target_modules=["q_proj","v_proj"])
            model = get_peft_model(model, lora_cfg)
            

            trainer = SFTTrainer(model=model, train_dataset=ds, ...)
            trainer.train()
            

            Step 500/500  loss: 0.8312
             LoRA adapter merged & saved
             Model pushed to HuggingFace Hub
            

            $ python evaluate_safety.py 
          

Download Curriculum

Choose your preferred internship duration and download the detailed curriculum to plan your learning journey

What You'll Learn

Six Core Skill Domains

A curriculum shaped by practising AI engineers and researchers building with frontier models at product companies and labs.

🧠

Prompt Engineering & Chain Design

Master zero-shot, few-shot, chain-of-thought, and tool-use prompting. Build reliable prompt chains with LangChain and LlamaIndex for real application workflows.

LangChainLlamaIndexOpenAI API

⚙️

Fine-Tuning & PEFT

Fine-tune open-source LLMs (Mistral, LLaMA) on domain datasets using LoRA and QLoRA. Manage GPU memory efficiently with 4-bit quantisation and evaluate with BLEU and perplexity.

LoRA / QLoRAPEFTTRL

🎨

Image Generation & Diffusion Models

Build text-to-image and image-to-image pipelines with Stable Diffusion. Apply ControlNet, DreamBooth, and SDXL for style-consistent and domain-specific generation tasks.

Stable DiffusionControlNetDiffusers

🔍

RAG & Knowledge Retrieval

Build retrieval-augmented generation systems using vector databases. Embed, index, and query large document corpora to ground LLM outputs in factual, up-to-date context.

PineconeChromaFAISS

🛡️

Safety, Bias & Evaluation

Identify and mitigate hallucination, toxicity, and bias in LLM outputs. Implement RLHF concepts, red-teaming protocols, and automated evaluation pipelines using LLM-as-judge patterns.

TruthfulQARLHFGuardrails

🚀

Production Integration & APIs

Wrap fine-tuned models and RAG pipelines as FastAPI services. Deploy to cloud with streaming inference, rate limiting, caching, and observability using LangSmith and Weights & Biases.

FastAPILangSmithW&B

Program Timeline

Your Journey, Month by Month

A structured ramp from LLM fundamentals to shipping production-ready Generative AI applications.

MONTH 1

LLM Foundations & Prompt Engineering

Transformer architecture deep-dive, attention mechanisms, and tokenisation. Master zero-shot, few-shot, chain-of-thought, and tool-use prompting patterns. Build your first LangChain application using OpenAI and open-source model APIs. Mentorship kick-off with your assigned AI engineer.

MONTH 2

Fine-Tuning, RAG & Image Generation

Fine-tune Mistral or LLaMA on a domain dataset using LoRA/QLoRA. Build a full RAG pipeline with Chroma and LlamaIndex. Simultaneously explore Stable Diffusion pipelines ? text-to-image, ControlNet, and DreamBooth personalisation.

MONTH 3

Safety, Evaluation & Alignment

Implement red-teaming protocols, evaluate model outputs for hallucination and toxicity, and apply guardrail frameworks. Study RLHF concepts and preference datasets. Build automated LLM-as-judge evaluation pipelines tracked in Weights & Biases.

MONTH 4 / GRADUATION

Production Deployment & Capstone

Package your Gen AI system as a FastAPI service with streaming inference, deploy to cloud, and connect a Streamlit or Gradio frontend. Present your capstone application to industry guests and receive your verified certificate, LinkedIn endorsement, and referrals to AI-first hiring partners.

Tech Stack

Tools You'll Master

Python 3.12

HuggingFace Transformers

PEFT / LoRA / QLoRA

LangChain

LlamaIndex

Stable Diffusion / SDXL

Diffusers / ControlNet

Pinecone / Chroma / FAISS

FastAPI

Weights & Biases

LangSmith

Gradio / Streamlit

AWS / GCP

Guardrails AI

Eligibility

Who Should Apply?

We value Python fluency and ML curiosity over existing LLM credentials ? the field moves fast and we teach what matters now.

Ideal Candidates

CS, IT, maths, or linguistics students (bachelor/master)
Solid Python - functions, classes, API calls, async
Basic ML understanding - supervised learning, loss functions
Experimented with ChatGPT, Claude, or open-source LLMs
Completed at least one ML or NLP project or course
Fascinated by how language models reason and generate

Common Barriers (We Help With)

No prior LLM or Gen AI experience required
No deep learning or PyTorch background needed upfront
No research publications or certifications mandatory
Non-CS backgrounds (linguistics, philosophy, design) welcome
Part-time track available for working students

Application

Start Your Application

Application Submitted!

Thank you! We've sent a confirmation to your inbox.
Our team will reach out within 2?3 business days.

Full Name

Email Address

Phone Number

Current Qualification

Preferred Track

Select Internship Domain

Start Availability

Why are you right for this internship?

Your information is encrypted and never shared with third parties.

FAQ

Common Questions

Is this internship paid?

Stipends for outstanding performers from month 2. All interns receive a verified certificate, LinkedIn endorsement, and placement support at AI-first startups, research labs, and product companies.

Can I do this while studying full-time?

Yes ? our part-time track requires around 20 hrs/week, is structured around academic schedules, and includes flexible lab windows and recorded sessions for async work.

What equipment do I need?

A modern laptop (8 GB+ RAM) and stable internet. GPU-intensive fine-tuning runs on cloud compute we provide ? no expensive local hardware required.

How competitive is selection?

We accept roughly 20% of applicants per cohort, prioritising Python fluency, ML fundamentals, and genuine curiosity about generative models over prior industry experience.

Will I work with real frontier models?

Yes ? interns get API access to OpenAI GPT-4o, Anthropic Claude, and open-source models like Mistral and LLaMA, working on real client and internal AI projects.

What career paths does this open?

AI engineer, LLM engineer, prompt engineer, MLOps engineer, NLP researcher, and AI product manager roles at startups, consultancies, research labs, and enterprise AI teams.

Prompt. Fine-Tune. Deploy.Generative AI Internship

Real Models. Real Pipelines.Real Products.