LLM Integrations Engineer

I turn frontier models into production systems.

I design the harnesses, retrieval pipelines, and serving infrastructure that sit between a raw language model and a real product — the parts that decide whether an LLM feature is reliable, fast, and cheap, or none of those things.

01

Agentic Harnesses

Task-specific scaffolding that makes a model dependable at one job. The harness — not the model — is usually what's broken when an agent loops, hallucinates a tool call, or silently does the wrong thing.

🔧

Tool calling

Designing tool schemas, argument validation, and error-recovery loops so the model picks the right call and recovers gracefully when it doesn't.

Typed schemas, retry/repair on malformed args, parallel tool dispatch.
🧱

Structured outputs

Constraining generations to JSON / schema-conformant shapes that downstream code can trust, with graceful fallback when the model goes off-format.

JSON-schema constraints, partial-parse handling, validation gates.
🔁

Agentic workflows

Multi-step plan → act → observe loops with the right stopping conditions, budgets, and guardrails so agents finish the task instead of wandering.

State machines, step budgets, human-in-the-loop checkpoints.
function-calling json-schema ReAct tool-use guardrails
What you get

An agent that completes its task reliably, fails loudly instead of silently, and is cheap to extend with new tools.

02

Retrieval & Knowledge

Connecting models to your actual documents and data — so answers are grounded in your knowledge base instead of the model's training data. Covers ingestion, retrieval, and summarization end to end.

📚

Retrieval-augmented generation

Chunking, embedding, and retrieval strategy tuned to your corpus — plus the reranking and prompt assembly that decide whether RAG actually helps.

Hybrid search, reranking, context-window packing, citation grounding.
🗄️

Vector databases

Choosing, indexing, and operating the vector store — embedding refresh, metadata filtering, and keeping recall high as the corpus grows.

pgvector / Pinecone / Weaviate, metadata filters, index maintenance.
📁

Google Drive & MCP sources

Wiring live document sources into the model via MCP servers, so the knowledge base stays in sync with where your team actually works.

MCP connectors, incremental sync, permission-aware retrieval.
📝

Document summarization

Map-reduce and hierarchical summarization for long documents, with structured extraction and optional human-labeled annotation for quality.

Long-doc summarization, structured extraction, data annotation loops.
RAG embeddings vector-db MCP reranking chunking
What you get

Grounded, citeable answers from your own data — with a retrieval layer you can measure and tune, not a black box.

03

Streaming & Performance

Latency is a product feature. I make LLM responses feel instant through streaming architecture and targeted optimization of the moments users actually perceive.

Time-to-first-token

Cutting the gap between request and first visible token — the single biggest driver of perceived speed — through prompt caching, routing, and warm paths.

Prompt caching, speculative prefetch, model/route selection.
🌊

Streaming architecture

Token-by-token streaming through the full stack — server, transport, and UI — including partial tool-call and structured-output streaming.

SSE / WebSocket transport, backpressure, partial-parse UI.
📉

Response-speed optimization

End-to-end latency budgeting: finding where the seconds go across retrieval, generation, and post-processing, then closing the gaps that matter.

Latency tracing, parallelization, payload trimming, caching tiers.
TTFT streaming SSE prompt-caching latency
What you get

Responses that start instantly and stream smoothly — measured in real TTFT and p95 latency, not vibes.

04

Model Proxying & Infra

The platform layer every LLM feature should sit behind: one gateway that handles routing, cost, observability, and safety — so individual features don't each reinvent it.

📊

In-house metrics

First-class observability for every call — tokens, cost, latency, cache hits, error rates — feeding dashboards and alerts you actually own.

Per-request tracing, cost attribution, custom metric pipelines.
🔀

Model switching & cost optimization

Routing each request to the cheapest model that can do the job, with fallback chains and A/B routing — often the fastest way to cut spend without losing quality.

Tiered routing, fallback chains, quality-gated downgrades.
🛡️

PII redaction

Detecting and stripping sensitive data before it reaches a provider, with reversible tokenization where the workflow needs the original back.

Inline detection, reversible tokenization, audit logging.
gateway routing cost-opt observability PII fallbacks
What you get

One proxy that makes every downstream feature cheaper, observable, swappable, and safe by default.

05

Conversational Products

Where the previous five layers come together into something users talk to. Support chatbots are the showcase: retrieval, harness, streaming, and proxy in one product, judged on real outcomes.

💬

Response generation

Grounded, on-brand answers that cite sources, know when to escalate to a human, and stay within policy — built on the retrieval and harness layers.

Grounded answers, escalation logic, tone & policy control.
⏱️

Support-speed optimization

Cutting time-to-resolution and first-response time — combining streaming, caching, and smart routing so customers feel helped immediately.

First-response latency, deflection rate, resolution time.
chatbots support escalation grounding deflection
What you get

A support assistant that resolves more tickets faster — measured on deflection and resolution time, not message count.

06

Evals & Reliability

The discipline that holds all the above accountable. Without evals, every change to a prompt or model is a guess. I build the measurement that turns "seems better" into a number.

🎯

Eval development

Task-specific eval suites — golden sets, rubrics, and graders — so you can compare prompts and models on the criteria that matter to your product.

Golden datasets, rubric design, regression gates in CI.
⚖️

LLM-as-judge

Automated graders calibrated against human labels, so evaluation scales beyond what you can hand-score while staying trustworthy.

Judge calibration, agreement tracking, bias checks.
🏷️

Data annotation

Building the labeled datasets that everything else depends on — annotation guidelines, inter-rater agreement, and feedback loops from production.

Labeling guidelines, agreement metrics, production feedback.
evals llm-as-judge golden-sets regression annotation
What you get

Confidence to ship changes — every model or prompt swap backed by a number that moves the right direction.

Have an LLM feature that needs to be production-grade?

Whether it's an agent that won't stay on task, a RAG pipeline that hallucinates, or a model bill that won't stop growing — I can help.

Get in touch →