Copy of Design & Discovery

Restaurant reservation and management platform

An AI-powered platform designed for restaurant discovery, data aggregation, and content enrichment. It leverages intelligent algorithms to unify information from various sources and generate engaging, context-aware content for enhanced user experience.

Our contribution

Typesense-powered vector search with support for filtering, sorting, and keyword constraints

Multilingual content classification, generation and translation using fine-tuned LLMs

Data aggregation and normalization across multiple sources

Core technologies

JS/React, Elixir, Postgress, Styled Components, GraphQl (Apollo), Rails3/5, Ruby 2.6, Redis, MongoDB, Kafka, Elasticsearch, Phoenix

Respondology

Respondology protects brands and individuals from the trolls and their toxicity on social media by removing hateful, ugly, and spam comments 24/7/365 using AI keyword filtering and human moderators.

Our contribution

Using Natural Language Processing and Machine Learning we helped to build a product to automatically moderate social media comments at scale, far faster and more consistently than a human team could.

Core technologies

Ruby/Rails, PostgreSQL, AWS

Main

NY Healthcare startup

A proprietary AI solution for speech recognition tailored to streamline digital documentation processes. Optimized for healthcare, it enables professionals to dictate notes efficiently, reducing time spent on manual data entry and improving workflow productivity.

Our contribution

AI-powered speech recognition technology that enables caregivers to dictate their notes into a smartphone.

AI speech processing, structures the content, and automatically generates the relevant documentation, such as patient reports, vital signs, and medication logs, directly into the system.

Core technologies

Ruby/Rails,React.js, PostgreSQL, Python, Whisper, vLLM and local Nvidia instruct model for HIPPA compliance

AINekko

A fully open-source company committed to democratizing collaboration across the AI community. Focused on models, datasets, and applications, it aims to ensure a diverse and open ecosystem, offering alternatives beyond closed, centralized platforms.

Our contribution

End-to-end testing and benchmarking of local LLM infrastructure, including token generation speed and latency analysis
OpenAI-compatible API layer for llama.cpp with support for advanced features like function calling, vision, and multi-model routing
Developer-focused reference implementations and tooling for integrating local models into low-level systems and applications
Tools for building open datasets and enabling modular AI system experimentation within open-source environments

Core technologies

Ruby/Rails, React.js, PostgreSQL, Python

Talentako

An application designed for hiring agencies to streamline the anonymization of candidate CVs. It efficiently masks sensitive information, enabling safe sharing with clients while reducing the risk of direct hiring and protecting candidate privacy.

Our contribution

Developed an MVP tailored for recruitment workflows, with built-in tools for anonymizing candidate CVs

Integrated automated detection and highlighting of sensitive information using a custom neural network

Enabled faster manual review and redaction through smart pre-selection of data points

Designed a user-friendly interface for hiring agencies to efficiently process and share anonymized CVs with clients

Core technologies

Ruby/Rails,React.js, PostgreSQL, Python

Llamagator

An LLM aggregator designed as a multi-LLM prompt testing tool. It enables rapid comparison, evaluation, and optimization of prompts across different large language models within a unified interface.

Our contribution

Test prompts against multiple LLMs or LLM versions
Observe the relative performance of generated responses
Run prompts multiple times to observe the reliability
Supports local and API access to LLMs
Licensed under Apache License, Version 2.0

Core technologies

Ollama, llama.cpp, OpenAI API

Blame.ai

A demo application built to test AI infrastructure components, from large language models to inference engines. It provides a controlled environment for validating performance, compatibility, and deployment workflows.

Our contribution

Building an open dataset

Integration with low level application code

Reference implementation for developers showcase based on local AI infrastructure.

Core technologies

Ruby/Rails,React.js, PostgreSQL, Python

NekkoAPI

A web API server compatible with OpenAI’s API specifications, built to interface with llama.cpp. It enables seamless integration of local LLMs into existing applications using familiar API calls.

Our contribution

Low-level access to C API via ctypes interface.
High-level Python API for text completion
OpenAI-like API LangChain compatibility
LlamaIndex compatibility
OpenAI compatible web server
Function Calling support
Vision API support
Multiple Models

Core technologies

Ruby/Rails,React.js, PostgreSQL, Python

Turtlenekko

A Go-based benchmarking tool for evaluating the performance of local large language models via chat completion endpoints. It delivers detailed analysis of token generation speeds and overall responsiveness, supporting optimization and performance tuning.