ALLO NUTRI FEED FORMULATOR

Inspiration

I grew up in a farming family in northern Benin, where poultry production is an important source of food and income. Later, while working as a veterinarian with farms raising approximately 3,000 broilers, layers, and guinea fowl, I witnessed how climate variability directly affected feed availability and prices. During periods of drought and irregular rainfall, the prices of major ingredients such as maize and soybean increased significantly. Farmers were forced to replace them with locally available alternatives, often without knowing whether the new formulation still met the nutritional requirements of their birds. Poor substitutions resulted in slower growth, reduced egg production, higher mortality, and financial losses. Most smallholder farmers cannot afford professional feed-formulation software or regular support from an animal nutritionist. Existing tools are often complex, expensive, internet-dependent, or poorly adapted to locally available African feed ingredients. This experience inspired me to build ALLO Nutri Feed Formulator: an offline, AI-assisted decision-support tool that helps smallholder poultry farmers formulate affordable, nutritionally balanced, and climate-resilient feeds using ingredients available in their communities.

What it does

ALLO Nutri Feed Formulator helps poultry farmers create affordable, balanced, and safe feed using locally available ingredients. The farmer selects the production phase, enters the desired feed quantity, chooses available ingredients, and provides their current prices.

The system then calculates the least-cost formula while checking the birds’ nutritional requirements and safe ingredient-inclusion limits. Within seconds, it displays the quantity of each ingredient, the total cost, the nutritional composition, and any safety alerts. If a balanced formula cannot be generated, it explains the problem and suggests possible corrections or local alternatives. Its offline-oriented design makes it suitable for rural areas with limited internet access.

How we built it

The project combines poultry nutrition, mathematical optimization, local ingredient data, and a lightweight language model.

We first created a technical database containing:

• the nutrient composition of locally available ingredients;
• the nutritional requirements of broiler chickens at different production stages;
• recommended ingredient-inclusion limits;
• antinutritional factors;
• safety rules; and
• vitamin and mineral requirements. The formulation engine compares the selected ingredients with the nutritional requirements of the birds and searches for a feasible least-cost combination. It also returns specific messages when:
• fewer than two ingredients are selected;
• a required premix is missing;
• the minimum inclusion level exceeds the maximum or no feasible formulation can be generated from the available ingredients.

For the AI component, we integrated a locally deployable quantized language model based on Qwen2.5-1.5B. The model is designed to explain formulation results, nutritional deficiencies, safety alerts, and possible ingredient substitutions in simple language. A commercial spreadsheet-based interface was also developed to make the system familiar and accessible to farmers, technicians, cooperatives, and agricultural-support organizations. It includes controlled ingredient selection, nutritional indicators, formulation results, inclusion-limit alerts, and a licensing system.

Challenges we ran into

One of our main challenges was adapting the tool to farmers’ actual feed-formulation practices. Testing the spreadsheet with 11 farmers showed that they use different ingredients, measurement units, production targets, and purchasing methods. We therefore had to make the interface flexible while keeping it simple for users with different levels of digital literacy. Another challenge was building a locally relevant ingredient database, since the nutritional composition of feed ingredients can vary depending on their variety, origin, processing method, season, and storage conditions. Although the prototype currently uses established technical reference values, laboratory analyses will be required to calibrate the database using ingredients collected in Benin.

Generating feasible, affordable, and safe formulations was also challenging. Some combinations of ingredients selected by farmers cannot meet all the nutritional requirements of the birds, while the least expensive combination may exceed safe inclusion levels or create nutritional deficiencies. The system must therefore prioritize animal health and performance and clearly explain why a feasible formulation cannot sometimes be generated. In addition, the farmers’ feedback showed that users need more than nutrient values and percentages. They also need simple explanations of deficiencies, ingredient limits, possible substitutions, and corrective actions. Translating complex animal-nutrition science into practical recommendations remains a major design challenge.

Accomplishments that we are proud of

We successfully developed a functional spreadsheet-based prototype that combines least-cost feed formulation, nutritional requirements, ingredient-inclusion limits, antinutritional-factor alerts, and AI-supported explanations. Most importantly, we tested the spreadsheet with 11 poultry farmers. They appreciated the tool, particularly its ability to use locally available ingredients, consider current market prices, simplify feed formulation, identify nutritional imbalances, and prevent unsafe ingredient-inclusion levels. Their positive feedback confirmed that ALLO Nutri addresses a real need and provided valuable insights into usability, ingredient availability, language preferences, and the types of explanations farmers require. We are especially proud that ALLO Nutri brings together poultry-nutrition science, mathematical optimization, locally relevant ingredients, safety controls, a familiar spreadsheet interface, lightweight generative AI, and offline deployment potential. This initial field test demonstrates that ALLO Nutri is more than a technical concept: it is a practical solution that has already been used and appreciated by its intended users.

What we learned

Building ALLO Nutri taught us that feed formulation is not only an optimization problem but also a matter of accessibility, trust, local relevance, and risk communication. We learned that the least expensive ration is not necessarily the safest or most productive, and that local feed resources can strengthen climate resilience when used at appropriate levels. We also discovered how an optimization engine and a language model can complement each other, particularly through offline AI in communities with limited connectivity. Most importantly, users need clear explanations rather than numerical results alone, while field testing is essential for improving both technical recommendations and user experience. This project helped us integrate animal nutrition expertise with lightweight generative AI and design a solution adapted to the practical realities of its intended users.

What's next for ALLO NUTRI FEED FORMULATOR

What is next for ALLO Nutri?

The feedback received from the 11 farmers will guide the next phase of development. We will improve the spreadsheet interface based on their experience and expand testing to a larger and more diverse group of poultry farmers. During this next pilot, we plan to compare ALLO Nutri formulations with the farmers’ existing formulas and measure differences in feed costs, bird growth, feed conversion, mortality, egg production, and profitability. We will also conduct laboratory analyses of locally sourced feed ingredients and validate selected formulations through controlled feeding trials.

The next version will include more local and climate-resilient ingredients, stronger alerts for antinutritional factors and unsafe inclusion levels, and formulations for layers, guinea fowl, and other poultry categories. We also plan to provide recommendations in French and relevant local languages, improve the offline AI assistant, simplify its explanations, and develop a mobile-friendly version for farmers who do not have regular access to computers. In the longer term, the platform will incorporate seasonal price and climate data to anticipate ingredient shortages and recommend suitable alternatives before farmers experience major disruptions.

Our immediate market objective is to grow from the initial 11-farmer pilot to 200 farmers during the first year. We then aim to reach 1,000 farmers across Benin within three years and 5,000 farmers in West Africa by 2030. Ultimately, ALLO Nutri will become a comprehensive, locally adapted decision-support platform that helps poultry farmers reduce feed costs, respond to climate-related ingredient shortages, improve productivity, and protect animal health.

Built With

• ai
• css
• excel
• gguf
• github
• html
• javascript
• linear
• llama.cpp
• mathematical
• microsoft
• optimizely
• phyton
• programming
• qwen
• render
• ubuntu
• vba