SankalpRaiGambhir

AI Engineering

Reliable AI features: structured outputs, retrieval, streaming, evaluation, and cost-aware operation.

LLM orchestrationRAGOpenAI APIHugging FaceLangChainLangGraphLlamaIndexLangSmithLangfuseKnowledge graphsModel Context Protocol (MCP)AutoGenCrewAIQdrantVector searchPrompt & schema designSSE streamingCaching & cost control

AI earns its place when it removes costly manual work, shortens a decision cycle, or makes a product materially more useful—not when it produces an impressive demo. I focus on workflows with measurable leverage: reliable extraction, assisted decisions, knowledge retrieval, and bounded automation that a team can operate and improve.

The scars are predictable: an agent loops until the bill arrives, retrieval returns an obsolete document with total confidence, a model emits output the next service cannot parse, or a harmless prompt change doubles cost. I build in schemas, state boundaries, evaluation sets, tracing, retrieval filters, caching, rate limits, and cost budgets from the start—so the AI feature has a clear failure mode, an owner, and a defensible return on spend.

Resume Diff AI dashboard showing AI-generated resume-to-job-description match scoring

Automating Resume Screening with Reliable Structured Outputs from LLMs

Screening a few hundred resumes eats days of recruiter time and still ends in gut calls. This scores every candidate in seconds, consistently enough to defend to a hiring manager.

View Technical Details

Document management RAG chat interface showing semantic search results with source attribution

Scaling AI Q&A Over Documents at Low Cost with Cached Embeddings & Vector Search

The answer is in the documents somewhere — finding it costs hours. Ask in plain English, get a sourced answer in seconds, on infrastructure that costs almost nothing to run.

View Technical Details

Agent control-flow diagram showing a runaway tool-call loop bounded by circuit breakers

Our AI Agent Made 2,000 Tool Calls to Do Nothing

What an autonomous agent taught me about cost, control flow, and why reliability isn't a model-intelligence problem.

Read Insight

LLM cost optimization dashboard showing token usage and spend trending down

Our LLM Bill Hit $14K a Month. I Cut It 70% — Without Cutting Quality

How instrumenting LLM spend, fixing prompt bloat, model routing, and safe caching cut a $14K/month AI bill by 70% without hurting quality.

Read Insight

RAG retrieval pipeline diagram showing chunking, vector search, and reranking stages

Our RAG Chatbot Answered Confidently — From the Wrong Document

How scaling decisions around chunking, vector search, quantization, and reranking quietly made our production RAG system less accurate.

Read Insight

Need AI Engineering Work Done?

Grab a slot below for a 30-minute call.

This site uses analytics cookies to understand how visitors use it. See the Privacy Policy for details.

AI Engineering

Relevant Selected Work

Automating Resume Screening with Reliable Structured Outputs from LLMs

Scaling AI Q&A Over Documents at Low Cost with Cached Embeddings & Vector Search

Relevant Engineering Insights

Our AI Agent Made 2,000 Tool Calls to Do Nothing

Our LLM Bill Hit $14K a Month. I Cut It 70% — Without Cutting Quality

Our RAG Chatbot Answered Confidently — From the Wrong Document

Need AI Engineering Work Done?

AI Engineering

Relevant Selected Work

Automating Resume Screening with Reliable Structured Outputs from LLMs

Scaling AI Q&A Over Documents at Low Cost with Cached Embeddings & Vector Search

Relevant Engineering Insights

Our AI Agent Made 2,000 Tool Calls to Do Nothing

Our LLM Bill Hit $14K a Month. I Cut It 70% — Without Cutting Quality

Our RAG Chatbot Answered Confidently — From the Wrong Document

Need AI Engineering Work Done?