AiGanakAkshay Bankar
Enterprise AI architect and consultant. Small language models, secure on-premise AI and agentic systems, built on 16 years of software architecture.
Akshay Bankar is the founder and director of AiGanak, an enterprise AI architect with more than 16 years in software architecture, specialising in small language models, secure on-premise AI and agentic systems.
Architecting efficient and secure AI
For more than 16 years I have turned complex technical problems into high-performance enterprise systems, moving from distributed systems to deep learning, generative AI and small language models.
I bridge research and market-ready delivery: systems that optimise AI rather than only call it, cutting operating cost and protecting data through fine-tuned models and secure RAG.
SLMs and RAG
Optimising 1B-parameter models for 128k context while keeping logic and structure, at far lower token cost than large models.
Cloud-native architecture
AWS, Java, Python and distributed systems for AI pipelines at enterprise load.
Secure on-premise AI
Agents that reason and run workflows inside a company's own infrastructure.
Distributed teams
Clear technical guidance, mentoring and lean delivery with remote international teams.
How I can help your team
AI architecture review
An independent review of your AI stack, cost, security and scaling risks, with a written plan.
SLM and RAG strategy
When a small fine-tuned model beats a large API, and how to build the retrieval layer.
Private and on-prem AI
Design for air-gapped or local-first AI where data privacy is critical.
Agentic systems
Agents that reason and execute workflows, with orchestration, tracing and guardrails.
AI cost optimisation
Find token waste and technical debt and cut operating cost without losing quality.
Fractional AI architect
Ongoing technical leadership for teams shipping their first production AI.
Recent projects
AiGanak-SLM-1B
A 1B decoder for logistics data with 128k context, 4-bit quantization-aware training and attention steering for IDs and timestamps.
AiGanak-AIOps-MCP
Reference architecture for autonomous AI engineering agents: FastMCP, a LangGraph cyclic orchestrator, DuckDB log analytics and Arize Phoenix OpenTelemetry tracing.
Vectomat Insights
The AI layer that lets expert-validated market research scale across industries and geographies.
Career
- Apr 2021 – presentFounder & Director · Technical Director, AI SolutionsAiGanak
- Jun 2018 – Mar 2021Software ArchitectCrossover for Work
- Apr 2016 – Jun 2018Software Developer IIOptiva Inc. · Pune
- Mar 2015 – Mar 2016Consultant, TechnologyDeloitte
- Mar 2010 – Mar 2015Senior Software EngineerTech Mahindra
- AI Agents and Agentic AI with Python & Generative AI
Vanderbilt University · Apr 2025 - Agentic AI and AI Agents: A Primer for Leaders
Vanderbilt University · Apr 2025 - Generative AI with Large Language Models
DeepLearning.AI & AWS · Mar 2025
BITS Pilani
Tell me what you're building.
I reply within one working day with next steps, or book a 30-minute call directly.
Questions people ask
Can't find it? Press ⌘K and ask.
Who founded AiGanak?
Akshay Bankar, an enterprise AI architect with more than 16 years in software architecture. He leads AiGanak's AI solutions and its small language model research.
Does Akshay Bankar consult independently?
Yes, through AiGanak: architecture reviews, SLM and RAG strategy, private AI design, agentic systems, AI cost optimisation and fractional AI architect roles.