Skip to content
TechMonster™

AI-Native / LLM Application Engineering

Build production LLM applications the way modern teams actually do: structured outputs, RAG, agents, tool use, evals, and MCP.

Build production LLM applications the way modern teams actually do: structured outputs, RAG, agents, tool use, evals, and MCP. Every project demands you explain the model's behavior, not just call an API.

4 phases · 17 projects · v2.1.0

Target roles: AI Engineer, LLM Application Developer, Applied AI Engineer, ML-adjacent Software Engineer

Stack: Python, TypeScript, LLM provider SDK, FastAPI/Next.js, pgvector, Docker

PHASE 1 · 4 projects

LLM Foundations & Structured Output

Call models reliably, get structured data out, and reason about cost and failure.

  1. 01 · featured

    Prompt Engineering & Structured Output

    Typed LLM client that returns JSON validated against a schema, with retries

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  2. 02 · featured

    System Vs User Prompts & Roles

    A persona-configurable assistant where the system prompt is data, not hard-coded

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  3. 03 · featured

    Tokens, Context Windows & Cost

    A token + cost estimator for a conversation, with a context-window budget guard

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  4. 04 · featured

    Prompt Caching

    Add prompt caching to the assistant and measure the cache hit rate + cost delta

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

PHASE 2 · 5 projects

Retrieval-Augmented Generation

Ground model answers in your own data and explain why retrieval helps.

  1. 01 · featured

    Embeddings & Vector Similarity

    Semantic search over a document set using cosine similarity

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  2. 02 · featured

    Chunking Strategies

    A chunker that splits docs sensibly, with a test showing retrieval-quality differences

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  3. 03 · featured

    Vector Store (pgvector)

    Persist embeddings in pgvector and query top-k by similarity

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  4. 04 · featured

    RAG Pipeline End-to-end

    A 'chat with your docs' app: retrieve, inject context, and cite sources

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  5. 05 · featured

    Retrieval Evaluation

    Measure retrieval precision/recall on a small labeled query set

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

PHASE 3 · 4 projects

Tool Use & Agents

Let models take actions safely and orchestrate multi-step work.

  1. 01 · featured

    Tool Use / Function Calling

    An assistant that calls 2-3 real tools (e.g. calculator, web lookup, DB query)

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  2. 02 · featured

    Agent Loops & Stopping Conditions

    A multi-step agent that plans, acts, and knows when to stop

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  3. 03 · featured

    MCP (Model Context Protocol)

    Expose a small toolset as an MCP server and call it from a client

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  4. 04 · featured

    Memory & State Across Turns

    An agent that maintains and summarizes state across a long conversation

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

PHASE 4 · 4 projects

Evals, Safety & Production

Ship an LLM feature you can trust, measure, and operate.

  1. 01 · featured

    Evaluations (evals)

    An eval suite that scores your app's outputs against a rubric or golden set

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  2. 02 · featured

    Guardrails & Prompt Injection

    Defend the RAG/agent app against a prompt-injection attempt and document the mitigation

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  3. 03 · featured

    Streaming & UX

    Stream tokens to a UI with graceful cancellation

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

  4. 04 · featured

    Deployment & Observability

    Deploy one LLM app with request logging, cost tracking, and error handling

    Python · TypeScript · LLM provider SDK · FastAPI/Next.js · pgvector · Docker

Next step

Request AI-Native / LLM Application Engineering

An advisor reviews the fit and confirms terms in writing. A request is not an acceptance and nothing is charged until you agree.