Vinay Krishnan
Vinay Krishnan
Lead AI/ML Engineer  ·  AI Consultant  ·  14+ Years Experience

I architect and build production-grade AI systems — from agentic automation layers and conversational chatbots to on-device voice interfaces and LLM-powered developer tools. I specialize in decomposing complex, ambiguous problems into well-scoped, executable components, and in keeping AI systems lean, maintainable, and free of unnecessary framework bloat. I've collaborated with teams across India, the USA, Germany, and Israel.

14+
ASR
LLM
TTS
Python

Services

AI Transformation

Identifying high-value AI use cases for your organization and translating them into concrete project plans — from feasibility through roadmap.

Agentic Digital Coordination

Designing and building a highly connected agentic automation layer that orchestrates your existing digital systems, data sources, and APIs into coherent workflows.

Voice Interfaces

Handsfree, speech-first interfaces — wake-word detection, ASR, TTS, and LLM logic — orchestrated into your existing products and backend systems.

De-bloating

Rescuing agentic projects from heavy frameworks like LangChain or LlamaIndex. Replace opaque abstractions with lean, auditable code that you actually own.

Code Modernization

Using agentic tools to systematically refactor and modernize legacy codebases — adding tests, documentation, and structure at scale.

Hiring & Training

Helping teams plan AI headcount, consolidate requirements, conduct technical interviews, and build structured onboarding and training programs for AI roles.

Products

ZenoCortex

A prompt-driven, hyper-extensible chatbot core built for production. Instead of hardcoding agent logic in Python, ZenoCortex uses a table-of-contents-driven RAG architecture to dynamically select instructions at runtime — making it trivial to add new capabilities without cascading code changes.

Available as Python source code for licensing. Suitable for teams that want a solid, maintainable chatbot foundation without framework lock-in.

Contact for licensing →

Open Source

GitHub: github.com/charstorm

LLMBinge

Generative AI tool to generate and explore Wiki-like articles using LLMs. Great for deep-diving into a topic through an auto-generated knowledge graph.

View on GitHub →

LinkItAll

A dependency-graph tool for capturing hierarchical knowledge structure — ideas defined in YAML with prerequisite dependencies, enabling comprehensive learning paths.

GitHub  ·  Demo →

Reshka

Handsfree transcription — a browser-based voice-to-text tool running client-side.

Live Demo →

Vilberta

An interactive voice assistant: three-stage ASR + LLM + TTS pipeline with intelligent interruption handling and MCP tool-calling support.

View on GitHub →

Experience

Jan 2024 – Jan 2026
Lead AI/ML Engineer
UST — Trivandrum, Kerala
Led generative AI initiatives across automotive and enterprise domains. Built HR chatbots, analytics chatbots with NL-to-SQL, a voice-controlled vehicle assistant, and API test case generation systems. Reduced Azure AI Search costs significantly through resource auditing. Managed 3 direct reports; architected the ZenoCortex chatbot core framework.
Jul 2022 – Dec 2023
Freelance AI/ML Engineer
Independent — Remote (international clients)
Worked with international clients including a 10-month engagement with Hi.Auto (Israel) on speech recognition and deep learning for trend prediction and time series analysis.
Jan 2020 – Sep 2022
Founder / ML Engineer
Nooromtech Private Limited — Kochi, Kerala
Founded a company building ultra-lightweight command-based ASR (~2 MB) deployable via WebAssembly in browsers. Developed C++ speech decoder, DNN acoustic model in PyTorch, ported to Linux, Windows, Android, iOS, Raspberry Pi, and Web.
Sep 2014 – Dec 2019
Speech Recognition Research Engineer
HSA Labs / Fraunhofer IDMT — Oldenburg, Germany
Research in large-vocabulary ASR, keyword spotting, and speech activity detection. Implemented TDNN models with float32→int8 quantization and ARM NEON optimization for embedded deployment.
Jun 2011 – Aug 2014
Earlier Roles
Qualcomm India · IIT Madras · Audience India
DSP engineering and SIMD optimization (Audience India); ASR research for government project (IIT Madras); Femtocell SON algorithms in embedded C++ (Qualcomm, Hyderabad).

Skills

AI / LLM
OpenAI Claude / Anthropic Azure OpenAI Amazon Bedrock OpenRouter vLLM llama.cpp RAG Prompt Engineering Fine-tuning (Unsloth / PEFT) MCP
Machine Learning
PyTorch TensorFlow / TFLite Scikit-Learn ONNX HuggingFace Transformers MLFlow
Languages
Python Go C++ C Bash JavaScript
Backend / Infra
FastAPI Flask Django Docker Kubernetes Azure AWS GCP
Data / Databases
PostgreSQL Snowflake DuckDB MongoDB QDrant Weaviate Azure AI Search
Speech / Embedded
ASR (Kaldi) TTS Wake-Word Detection WebAssembly ARM NEON / SIMD Model Quantization

Contact