Sanket Tikhande
Taking select freelance AI projects

I build GenAI systems that actually ship.

AI/ML Sr. Engineering Manager, GenAI Architect and AI Consultant. 14+ years of engineering leadership, from RAG pipelines at sub-1s latency to agentic workflows in production.

View my workGitHub profile
Thane, Mumbai, IndiaRemote-friendly worldwide
Ask Sanket

How do you serve millions of documents in under a second?

Hybrid retrievalRe-rankingCitation trackingMulti-vector search
Azure AI Search + PGVectorP95 < 1s
Azure OpenAILangChainLangGraphPGVectorFAISSOpenSearchFastAPIRedisDockerGPT-4oLlamaIndexHaystackDSPyPython Azure OpenAILangChainLangGraphPGVectorFAISSOpenSearchFastAPIRedisDockerGPT-4oLlamaIndexHaystackDSPyPython
About

Architecture depth, with delivery discipline.

I build production-grade GenAI systems: RAG pipelines serving millions of docs at sub-1s latency, LangGraph agentic workflows, Azure OpenAI integrations, and multi-agent orchestration.

With 14+ years of engineering leadership across SaaS, InsurTech, AI, and hospitality, I bring both the architectural depth and the delivery discipline your project needs.

Currently leading a 12-member cross-functional team and taking select freelance clients for ambitious AI challenges.

Impact

Numbers that matter.

14+Years of engineering leadership
12–15Engineers led
< 1sP95 RAG latency
30–40%Support effort saved
4Countries onsite delivery
Selected work

Projects & featured work.

01

Enterprise RAG System

Enterprise-scale RAG pipeline serving millions of documents at sub-1s P95 latency. Hybrid retrieval with re-ranking, citation tracking, and multi-vector search using Azure AI Search + PGVector.

Azure OpenAILangChainPGVectorAzure AI SearchFastAPIDocker
02

Bank AI Multi-Vector Search Engine

Hybrid semantic search combining dense + sparse vectors with OpenSearch and FAISS. Sub-second retrieval across millions of records with custom re-ranking.

FAISSOpenSearchPGVectorPythonFastAPIRedis
03

Conversational AI Platform

Production conversational AI platform with WebSocket streaming, session management, and context-aware multi-turn dialogue. Integrated with Jio's service automation layer.

GPT-4oFastAPIWebSocketsPostgreSQLRedisAzure OpenAI
04 · Open source

Contributor @ Dynavec

Dynavec is a production-grade vector database and retrieval engine. I contributed 12+ features across quantization, hybrid search, RAG integrations, and embedding optimization.

  • Scalar Quantization & OPQ: INT8 compression for 4× memory reduction; OPQ rotation improved Recall@10 by +1.4pp at the same vector footprint.
  • HNSW Hot Tier & PQ Persistence: in-process ANN search for ultra-low latency; codebook serialization for production reloads.
  • Hybrid Search & Learned RRF Weights: fused graph+ANN retrieval via RRF; ML-trained BM25/vector blend weights, zero manual tuning.
  • Cross-encoder Reranker & Vectorized MMR: two-stage precision reranking and diverse result sets.
  • RAG Integrations: native connectors for Haystack, DSPy, and LlamaIndex; embedding cache to cut latency and API cost.
Vector QuantizationHNSWANN SearchRRFHybrid SearchCross-EncoderMMRLlamaIndexHaystackDSPyPython
Services

What I build.

RAG & Knowledge

Knowledge systems

  • Enterprise document intelligence
  • Vector search, hybrid retrieval
  • Re-ranking & citation pipelines
  • Sub-1s P95 at millions of docs
Agentic AI

Automation that works

  • LangGraph multi-agent systems
  • Tool-calling workflows
  • Autonomous support agents
  • 30–40% effort reduction in production
Architecture

AI advisory

  • Model selection & cost optimization
  • MLOps setup & LLM monitoring
  • Fractional CTO-style AI leadership
  • Cloud-native, production-ready
Career

Fourteen years, one direction.

2024 – Present

Sr. AI/ML Engineering Manager

Leading 12-member team · Enterprise RAG · Agentic AI · Conversational AI

2020 – 2023

Tech Lead

InsurTech & SaaS · LLM agents

2015 – 2020

Programmer Analyst

SaaS & Hospitality · Team scaling · Python & Ruby on Rails · Onsite: 🇮🇩 🇻🇳 🇭🇰 🇹🇭

2012 – 2015

Software Developer

Sodel Solutions Pvt Ltd

🇮🇩 IndonesiaC2L Biz Solutions · Analysis Programmer · InsurTech
🇻🇳 VietnamC2L Biz Solutions · Analysis Programmer · InsurTech
🇭🇰 Hong KongC2L Biz Solutions · Analysis Programmer · InsurTech
🇹🇭 ThailandC2L Biz Solutions · Analysis Programmer · InsurTech
Why work with me

Built for real delivery.

Production, not demosArchitecture to deployment with monitoring baked in.
Hands-on depthI write the code, tune the prompts, and design the infra, not just slide decks.
Open source contributorActive Dynavec contributor in vector quantization, ANN search, hybrid retrieval.
Cross-border deliveryOnsite in 4 countries, plus distributed team leadership.
People leadershipHired, mentored, and scaled teams of 12–15 engineers.
Business-drivenEvery system tied to measurable outcomes: latency, cost, effort saved.

Let's build something with AI.

Got a RAG system to architect, an AI agent to ship, or a team that needs GenAI leadership? I've done it at enterprise scale.

Connect on GitHub