AI/ML Engineer Intern – LLM Training & Agentic AI

Horrazon AI | Horrazon Intelligence Private Limited

- Internship Duration: 6 Months
- Employment Type: Full-time Internship
- Work Mode: Remote
- Department: Artificial Intelligence & Machine Learning
- Experience Level: Experienced Candidates Only
- Future Opportunity: Potential Full-Time Employment Based on Exceptional Performance
- Application Email: [email protected]

About Horrazon AI

Horrazon AI is building intelligent systems that transform complex business data into actionable decisions. Our focus is on developing advanced language models, agentic AI systems, and production-grade intelligence infrastructure that help businesses understand their data, identify opportunities, and act with greater precision.

At the core of our technology is Pogee, our LLM initiative, supported by an intelligent agent layer designed to process information, reason across multiple sources, and generate decision-oriented insights.

Our platform connects business intelligence with AI-driven execution, working across data from platforms such as Instagram, Facebook, LinkedIn, YouTube, and X.

Horrazon is supported by startup and technology ecosystems, including NVIDIA Inception, AWS Startup, and Zoho for Startups. We operate with a lean engineering culture where technical ownership, measurable outcomes, and production quality matter more than titles or hierarchy.

We are building for real-world deployment—not demonstrations that never leave a notebook.

The Opportunity

We are seeking an experienced and technically strong AI/ML Engineer to work on LLM training and fine-tuning, model evaluation, and agentic AI engineering.

This opportunity is intended for graduates who already possess substantial hands-on experience building, training, fine-tuning, or evaluating machine learning and LLM systems.

We are not looking for someone whose experience is limited to prompt engineering, API integration, tutorials, or introductory AI projects. We expect applicants to understand the underlying technology, explain their engineering decisions, and demonstrate the ability to build reliable systems.

You do not need a fixed number of years in the industry. However, you must have meaningful practical experience and be prepared to demonstrate your technical depth through an engineering assessment and interview.

During the six-month engagement, you will work directly with the founding and engineering team on real product development, with responsibilities spanning experimentation, implementation, evaluation, optimization, and production deployment.

Key Eligibility Requirements

Please read before applying.

- Graduates only: Applicants must have completed their undergraduate degree or an eligible postgraduate degree in Computer Science, Artificial Intelligence, Machine Learning, Data Science, or a related technical discipline.
- Undergraduate students are not eligible: Candidates currently pursuing their bachelor's degree, including students seeking internships alongside their studies, should not apply.
- Experienced candidates only: This is not an entry-level training programme. Applicants must have demonstrable practical experience in LLMs, machine learning, or closely related engineering work.
- Full-time commitment is mandatory: Candidates must be available to work full-time for the entire six-month internship.
- Technical evidence is essential: Applicants must provide GitHub repositories, research work, model-training experiments, deployed projects, technical reports, or other credible evidence of their capabilities.

Applications that do not meet these minimum requirements may not be considered.

Key Responsibilities

- LLM Training and Fine-Tuning

- Fine-tune open-source language models using techniques such as LoRA, QLoRA, and parameter-efficient fine-tuning.
- Evaluate the suitability of full fine-tuning, supervised fine-tuning, and preference-optimization methods, including DPO and RLHF-style approaches, for specific product requirements.
- Develop reproducible training pipelines and experiment configurations.
- Diagnose training instability, overfitting, data-quality issues, and model-performance limitations.
- Make evidence-based decisions about model selection, adaptation, and deployment.

2. Transformer Architecture and Model Engineering
- Apply a strong understanding of transformer architectures, attention mechanisms, embeddings, tokenization, positional representations, loss functions, and optimization.
- Investigate model behaviour and performance at the architectural,



training, and inference levels.
- Understand the trade-offs between model capability, parameter count, context length, compute requirements, and deployment constraints.
- Contribute to model adaptation and, where justified by product requirements, pre-training or continued-pretraining experiments.

3. Agentic AI and Reasoning Systems
- Design and implement AI agents capable of multi-step reasoning, tool calling, structured information retrieval, and task execution.
- Build reliable orchestration pipelines that connect language models with APIs, databases, analytical tools, and business workflows.
- Develop mechanisms for context management, state handling, structured outputs, error recovery, and controlled execution.
- Contribute to multi-agent architectures where they provide measurable advantages over simpler alternatives.
- Help transform raw business and marketing data into reliable, explainable, actionable recommendations.

4. Model Evaluation and Experimentation
- Develop evaluation datasets, benchmarks, and task-specific testing frameworks.
- Measure model quality, factual accuracy, instruction adherence, latency, and operational reliability.
- Conduct controlled experiments and compare approaches using reproducible metrics.
- Identify failure modes, hallucinations, and edge cases before deployment.
- Document findings and make recommendations based on measurable results rather than subjective impressions.

5. Training Data and Data Engineering
- Prepare, clean, validate, transform, and structure datasets for training, fine-tuning, and evaluation.
- Develop data-processing pipelines using Python and SQL.
- Work with structured and unstructured data originating from business systems and social and advertising platforms.
- Apply appropriate deduplication, data-quality checks, filtering, and dataset-versioning practices.
- Maintain data integrity and follow responsible data-handling practices.

6. Inference Optimization and Production Engineering
- Optimize inference latency, throughput, memory consumption, and operating costs.
- Explore quantization, batching, caching, and other appropriate performance techniques.
- Integrate model-serving components with backend services and production applications.
- Write maintainable, tested, documented code and contribute to debugging, monitoring, and deployment workflows.
- Make practical engineering decisions that balance model quality with real-world infrastructure and budget constraints.

Required Technical Qualifications

Applicants should demonstrate strong practical proficiency in the following areas:

Machine Learning and LLMs

- Strong Python programming skills.
- Hands-on experience with PyTorch and Hugging Face Transformers.
- Practical experience with Hugging Face PEFT or comparable fine-tuning tooling.
- A demonstrated project involving actual model training, fine-tuning, or systematic evaluation.
- A solid understanding of transformer fundamentals, training objectives, loss functions, optimization, and model evaluation.

Agentic AI

- Practical experience building LLM-powered systems that go beyond basic chat interfaces.
- Familiarity with tool calling, multi-step workflows, retrieval, context management, and agent orchestration.
- Experience with LangChain, LlamaIndex, or custom orchestration implementations is useful, but architectural understanding matters more than framework familiarity.

Software Engineering

- Working knowledge of SQL and relational databases.
- Experience integrating REST APIs and handling structured data.
- Proficiency with Git, debugging, testing, and writing maintainable code.
- Ability to communicate technical decisions clearly and work independently in a remote-first environment.

Professional Expectations

- Take ownership of assigned engineering problems from investigation through implementation and evaluation.
- Communicate progress, blockers, and experimental findings clearly.
- Demonstrate intellectual honesty about model limitations and engineering trade-offs.
- Work with urgency while maintaining technical quality and reliability.

Preferred Qualifications The following will strengthen an application:





- Experience with distributed or multi-GPU training.
- Practical experience with model quantization and optimized inference.
- Experience with supervised fine-tuning, preference optimization, or reinforcement learning for language models.
- Contributions to open-source ML or LLM projects.
- Published research, competitive machine learning achievements, or technically substantial independent projects.
- Experience with PostgreSQL, Go, cloud infrastructure, or production model-serving systems.
- Familiarity with social media, marketing analytics, or advertising platform integrations.
- Experience deploying AI systems that have been used by actual users.

These are differentiators, not substitutes for the required hands-on experience.

What You Can Expect at Horrazon

Direct Technical Ownership

Work on substantive engineering problems with direct exposure to the founding and engineering team. Your contributions will be evaluated on their technical quality and measurable impact.

Real Product Engineering

Contribute to the development of models, agents, evaluation pipelines, and backend systems intended for real product use—not isolated academic exercises.

End-to-End Experience

Gain exposure to the full engineering lifecycle, from experimentation and dataset preparation to evaluation, integration, optimization, and deployment.

High-Accountability Environment

Work in a lean team that values independent thinking, rigorous experimentation, clear communication, and dependable execution.

Potential Full-Time Employment

Exceptional performers may be considered for a full-time position at the end of the six-month internship. Conversion is performance-based and is not guaranteed. Decisions will depend on demonstrated technical ability, ownership, work quality, business requirements, and the availability of a suitable role.

Selection Process

Our selection process is designed to assess demonstrated engineering ability rather than theoretical familiarity alone.

- Application Screening: Review of educational eligibility, relevant experience, resume, and technical evidence.
- Technical Assessment: Evaluation of practical ML/LLM engineering ability, which may include a model-training, fine-tuning, evaluation, or agentic AI task.
- Technical Interview: Discussion of architecture, implementation choices, experimental results, trade-offs, and lessons learned from previous projects.
- Final Discussion: Assessment of role alignment, availability, communication, ownership, and full-time commitment.
- Offer and Onboarding: Selected candidates receive the relevant internship terms and onboarding details.

Candidates should be prepared to explain their own contributions to every project they present.

How to Apply

Send your application to [email protected].

Email subject: AI/ML Engineer Intern – LLM Training & Agentic AI | [Full Name]

Please include

- Resume/CV: An updated PDF highlighting relevant technical experience.
- Technical portfolio: GitHub, Hugging Face, Kaggle, research publications, technical reports, or project demonstrations.
- Relevant experience: A brief summary of your strongest LLM or ML engineering project, including the problem, approach, tools, evaluation method, and measurable results where available.
- Motivation: Three to five sentences explaining your interest in LLM engineering and agentic AI, and why your experience is relevant to Horrazon.
- Eligibility and availability: Your completed degree, current location and time zone, earliest joining date, and confirmation of full-time availability for six months.

Applications without sufficient technical evidence may not progress to the assessment stage.

Final Note

Horrazon AI is looking for engineers who have already invested serious effort in understanding and building AI systems.

If you are a graduate with meaningful hands-on experience, can defend your technical decisions, and are ready to take responsibility for production-oriented engineering, we encourage you to apply.

If you are currently an undergraduate, a beginner looking for introductory training, or someone whose experience is limited to basic API wrappers and prompt-based prototypes, this position is not the right fit.

Build seriously. Experiment rigorously. Ship responsibly.

Horrazon Intelligence Private Limited

Horrazon AI — Engineering Intelligence for Real-World Decisions

Pay: From ₹6,000.00 per month

Benefits

- Flexible schedule
- Internet reimbursement
- Leave encashment
- Paid sick time
- Work from home

Work Location: Remote

📌 AI / ML Engineer Intern: (LLM & Agentic AI) (India)
🏢 Horrazon Intelligence Private
📍 India