AI Engineer.
AI Engineer at Flid AI, building LeapView — an open-source, agent-native BI platform powered by Go and DuckDB. I specialize in scalable backend systems, AI agent architectures, and analytics-as-code engines. Open to freelance & contract work.
Building agent-native intelligence
& high-performance Go infrastructure.
I'm Ganesh Kambli — an AI Engineer at Flid AI, currently building LeapView — an open-source, agent-native BI platform powered by Go and DuckDB. I specialize in building scalable backend systems, governed semantic data layers, analytics-as-code engines, and autonomous AI agents.
I'm a Computer Engineering graduate from Savitribai Phule Pune University, Pune. I've published 2 research papers on distributed honeypot architectures and AI-driven threat classification, won intercollegiate competitions, and actively contribute to open source as a Project Admin & Maintainer for ECSoC'26.
Beyond my full-time role, I'm open to freelance & paid contract work — especially projects involving agent systems, Go infrastructure, and analytics platforms.
Work & recognition.
- Building LeapView — an open-source, agent-native BI platform allowing teams to define analytics as code and query through interactive dashboards and AI agents.
- Designing governed semantic data layers, high-performance distributed query engines (Go + DuckDB), and autonomous AI agent architectures for analytics-as-code workflows.
- Eliminated manual attendance tracking by building an end-to-end Absence Detection System (Python + Google Sheets API + GitHub Actions), reducing reporting time by ~80% and human error to zero.
- Built a real-time Google Sheets monitoring pipeline that automated data reconciliation across internal workflows, replacing a manual, error-prone process.
- Coursework: DSA, DBMS, OS, Computer Networks, OOP, Software Engineering, Machine Learning.
- Applied knowledge through multiple deployed projects, 1 internship, and 2 intercollegiate competition wins.
Publications & Research
Things I've built.
Open-source, agent-native BI platform. Build governed semantic models and dashboards as code, review every change in Git, and explore trusted analytics through interactive dashboards and AI agents. Powered by Go and DuckDB for high-performance distributed query execution.
Engineered a 100% offline, agentic GraphRAG system using LangGraph and NetworkX to autonomously traverse function-call dependency graphs from local directories or GitHub URLs. Resolves natural-language queries via AST-based chunking and hybrid retrieval (ChromaDB + BM25) with Reciprocal Rank Fusion, running Qwen2.5-coder locally with exact file-path and line-number citations and a codebase synchronization watchdog daemon.
Cloud-native multi-protocol honeypot infrastructure featuring real-time attack telemetry and an LSTM-based threat classification engine. Backed by 2 published research papers on distributed honeypot architectures and AI-driven threat classification.
Deep semantic text similarity and originality scoring engine utilizing modern transformer embeddings and vector indexing. Performs context-aware plagiarism detection beyond surface-level text matching.
Architected a decoupled, multi-service ML platform integrating an inference pipeline, distributed weather caching via Redis, and real-time client updates using React.js and Python. Implemented statistical model drift monitoring and MLOps metrics tracking to maintain prediction reliability as live data distributions shifted.
Built a fully serverless, event-driven URL shortener using AWS Lambda, API Gateway, DynamoDB, S3, and CloudFront. Demonstrates scalable, cost-efficient, and highly available cloud-native backend design capable of handling high-throughput request rates.
Engineered a persistent virtual file system in C++17 implementing physical block storage simulation, POSIX permissions, and concurrent inode access via a thread-safe Readers-Writer lock model. Passed a comprehensive security audit hardening I/O operations, Argon2id-based authentication, and protected B-Tree operations.
Production-grade GPU acceleration project demonstrating advanced CUDA optimization techniques for Huffman Encoding. Achieved substantial throughput gains (45+ GB/s) over CPU-based frequency counting through massive parallelism on NVIDIA GPUs, featuring warp-level aggregation kernels and shared memory histograms.
Re-engineered a legacy Flask + raw HTML application into a modern multi-tier architecture featuring a concurrent FastAPI backend and a Vite React + TailwindCSS SPA frontend. Containerized the development and production cluster using Docker to streamline environment parity.
Built a mobile application in Android (Kotlin) that intervenes on distracting app usage and offers gamified wellness quests and AI-guided check-ins. Aims to reduce doomscrolling and improve focus through mindful tech habits and responsive wellness metrics.
Let's build
something together.
Open to freelance & contract work, and collaborations on agent-native systems, Go infrastructure, and analytics as code. I respond to all messages — drop me a line.