AI System for Software Testing and QA

Your coverage report shows 90% line coverage, but production bugs still slip through. Why? Because line coverage does not account for business scenarios, boundary values, and integration seams. An AI QA system solves this by analyzing the AST, requirements from Jira, and mutation testing. Result: boundary case coverage increases from 30% to 95%. For over 5 years, we have deployed such systems on 50+ projects — from startups to enterprise. Average savings on QA team after implementation are 40–60%, and the number of production incidents drops by 50–70% within the first three months. We can evaluate your project in one day — just contact us. Order a preliminary analysis and find out how many bugs are hiding in your code.

Problems we solve

• False sense of security. 100% line coverage does not guarantee all business scenarios are tested. AI finds logical gaps.
• Test fragility. Tests break during refactoring — AI self-healing adapts them to the new architecture.
• Integration blind spots. Unit tests miss errors at component seams. AI generates integration tests based on call graphs.
• Manual labor costs. Test writers spend up to 60% of time on routine — AI handles generation of basic and boundary cases.

Components of the AI Testing System

[Code Analysis]        [Requirement Analysis]
  AST parsing            NLP from Jira/Confluence
       ↓                        ↓
[Test Generation Engine]
  Unit | Integration | E2E | API
       ↓
[Test Prioritization]
  Change Impact Analysis → run needed tests, not all
       ↓
[Result Analysis]
  Failure Classification + Root Cause Suggestion
       ↓
[Coverage Intelligence]
  Semantic gaps in coverage

Each component is a separate microservice communicating via RabbitMQ. This allows independent scaling of generation and analysis.

How AI finds semantic gaps in coverage

Traditional coverage tools (Istanbul, JaCoCo) count lines. Problem: 100% line coverage does not mean all business scenarios are tested. Our SemanticCoverageAnalyzer uses GPT-4o at temperature 0.1 to detect gaps:

from langchain_openai import ChatOpenAI
import ast
import textwrap

class SemanticCoverageAnalyzer:
    """Analyzes semantic gaps in test coverage"""

    ANALYSIS_PROMPT = """Analyze the function and existing tests.
Identify which business scenarios and boundary conditions are NOT covered.

Function:
```python
{function_code}

Existing tests:

{existing_tests}

Identify uncovered scenarios:

1. Boundary values (empty string, None, 0, max int, negative)
2. Parameter combinations
3. Error scenarios (exceptions, invalid input)
4. Concurrent access (if applicable)
5. Business rules in conditions

For each: describe the scenario + why it is important + possible bug if not tested. Return JSON: {{gaps: [{{scenario, importance, potential_bug}}]}}"""

def __init__(self):
    self.llm = ChatOpenAI(model="gpt-4o", temperature=0.1)

def analyze_function_coverage(
    self,
    function_source: str,
    test_source: str
) -> list[dict]:
    result = self.llm.invoke(
        self.ANALYSIS_PROMPT.format(
            function_code=function_source,
            existing_tests=test_source
        )
    )
    import json
    return json.loads(result.content)["gaps"]

def extract_functions_from_module(self, source: str) -> list[dict]:
    """Extracts functions from a Python module via AST"""
    tree = ast.parse(source)
    functions = []
    for node in ast.walk(tree):
        if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
            func_source = ast.get_source_segment(source, node)
            complexity = self._calculate_cyclomatic_complexity(node)
            functions.append({
                "name": node.name,
                "source": func_source,
                "complexity": complexity,
                "line_start": node.lineno
            })
    return sorted(functions, key=lambda x: x["complexity"], reverse=True)

def _calculate_cyclomatic_complexity(self, node) -> int:
    """Cyclomatic complexity — priority for testing"""
    complexity = 1
    for child in ast.walk(node):
        if isinstance(child, (ast.If, ast.While, ast.For, ast.ExceptHandler,
                               ast.With, ast.Assert)):
            complexity += 1
        elif isinstance(child, ast.BoolOp):
            complexity += len(child.values) - 1
    return complexity

| Parameter | Line coverage | Semantic coverage |
|-----------|---------------|--------------------|
| What it measures | Executed lines | Covered business scenarios |
| Example miss | — | None in aggregation, empty list |
| Bug detection | Only in executed paths | All possible inputs |
| Average analysis time | Instant | Depends on LLM (2-5 sec per function) |

<cite>"After deploying the AI system, the client noted: we found 15 critical bugs that manual tests missed" — from client feedback</cite>

<details>
<summary>Example semantic gaps report</summary>
```json
{
  "gaps": [
    {"scenario": "Empty list in aggregation", "importance": "high", "potential_bug": "ZeroDivisionError"}
  ]
}

Test generator with mutation testing

class AITestGenerator:
    UNIT_TEST_PROMPT = """Generate pytest unit tests for the function.

Function:
{function_code}

Uncovered scenarios (focus on these):
{gaps}

Requirements:
- Use pytest + pytest-mock
- Parametrize with @pytest.mark.parametrize where applicable
- For each test: Arrange-Act-Assert
- Tests for boundary values
- Tests for erroneous input data
- Mock for external dependencies

Return only code, no explanations."""

    async def generate_unit_tests(
        self,
        function_source: str,
        gaps: list[dict]
    ) -> str:
        gaps_text = "\n".join([
            f"- {g['scenario']}: {g['importance']}"
            for g in gaps[:5]  # top 5 by importance
        ])

        result = await self.llm.ainvoke(
            self.UNIT_TEST_PROMPT.format(
                function_code=function_source,
                gaps_text=gaps_text
            )
        )
        return result.content

    async def run_mutation_testing(self, source_file: str, test_file: str) -> dict:
        """Runs mutation testing via mutmut"""
        import subprocess
        result = subprocess.run(
            ["mutmut", "run", f"--paths-to-mutate={source_file}",
             f"--tests-dir={test_file}"],
            capture_output=True, text=True
        )

        # Analyze surviving mutants (tests did not catch change)
        survived = self._parse_survived_mutants(result.stdout)
        if survived:
            additional_tests = await self._generate_for_mutants(survived, source_file)
            return {"survived_count": len(survived), "additional_tests": additional_tests}

        return {"survived_count": 0, "mutation_score": "100%"}

Mutation testing is the only way to check whether tests actually catch logic errors. Our AI doesn't just generate tests — it cyclically improves them until all mutants are killed. This guarantees that tests protect against real defects, not just formal line coverage.

How AI testing integrates into CI/CD

# .github/workflows/ai-qa.yml
name: AI QA Analysis

on:
  pull_request:
    types: [opened, synchronize]

jobs:
  ai-test-analysis:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0  # needed for diff

      - name: Analyze changed files
        run: |
          git diff origin/main...HEAD --name-only --diff-filter=AM | \
            grep "\.py$" > changed_files.txt

      - name: Run AI coverage analysis
        run: |
          python qa_system/analyze_coverage.py \
            --changed-files changed_files.txt \
            --generate-missing-tests \
            --output coverage_report.json

      - name: Comment PR with AI findings
        uses: actions/github-script@v7
        with:
          script: |
            const report = require('./coverage_report.json')
            const comment = formatReport(report)
            github.rest.issues.createComment({
              issue_number: context.issue.number,
              body: comment
            })

After integration, every PR receives an automatic comment listing new tests, mutation coverage level, and remaining gaps. The developer can accept changes or request additional tests — no manual test code review required.

What is included in the work

• Current coverage analysis. Run the semantic analyzer on the entire codebase, output a prioritized report.
• Test generation. Automatically create unit, integration, and E2E tests for each identified gap.
• Mutation testing. Cycle "generate — run — improve" until 100% mutation coverage is achieved.
• CI/CD setup. Integrate analysis into pipeline (GitHub Actions, GitLab CI, Jenkins) with PR commenting.
• Documentation. Document all generated tests and support methodology.
• Team training. Workshop on working with the AI system and writing custom rules for your business.
• 3 months of support. Unlimited consultations and adjustments for new code versions.

Timelines

Timelines may vary depending on codebase size and requirement complexity. We offer a free preliminary assessment of your project in one day. Order it, and we will provide a detailed implementation plan.

Why AI testing is more effective than manual

Over 5 years of experience in AI/ML and 50+ deployed QA systems. We guarantee that after implementation you will see a real reduction in production bugs, not just growing coverage metrics. Get a consultation: we will show how AI transforms your QA process.

LLM Development: Fine-Tuning, RAG, Agents, and Production Deployment

Using GPT‑4 or Claude 3.5 Sonnet through a public API is not a solution — it's just a tool. When the requirement is to "make it like ChatGPT, but on our data," there is a real engineering challenge behind it: from prompt engineering to training a 70B model on your own infrastructure. End-to-end LLM solution development is a complex stack, and we have been doing it for over 5 years. During this time, we have completed over 20 projects in generative AI: from RAG systems for legal departments to custom support agents. Where exactly your task falls depends on data, latency requirements, budget, and how critical confidentiality is.

A typical situation: the client has already tried ChatGPT, but results are unstable — sometimes accurate, sometimes hallucinating. Or they need integration into a corporate portal while complying with security policies. Let's break down each layer of the stack in detail — from RAG to production deployment.

Why Do RAG Systems Break and How to Fix It?

RAG (Retrieval-Augmented Generation) looks simple: find relevant documents, put them in context, get an answer. In practice, it fails in several places.

Chunking without overlap. Classic mistake: chunk_size=512, overlap=0. If the answer lies across two chunks, retrieval won't find either with sufficient confidence. Solution: overlap 15–25% of chunk_size, or better yet, sentence-aware splitting with spaCy or NLTK instead of naive character splitting.

Poor embedder. text-embedding-ada-002 is good for general use, but on legal or medical texts, specialized models like E5-large-v2, BGE-M3, or fine-tuned sentence-transformers on domain data outperform it. Recall@5 differences can be 15–25%.

No re-ranking. Vector search optimizes for speed, not relevance. A cross-encoder re-ranker (ms-marco-MiniLM-L-6-v2, bge-reranker-large) after initial retrieval improves top-3 accuracy with acceptable latency (+50–150ms). This is often more impactful than improving the embedding model.

Hybrid search. Dense vectors alone work poorly on exact queries: names, SKUs, codes. BM25 (sparse) finds exact matches but misses semantics. Hybrid via RRF (Reciprocal Rank Fusion) is the optimal compromise. Qdrant, Weaviate, and pgvector 0.7+ support hybrid search natively.

When to Fine-Tune Instead of Prompt Engineering?

Prompt engineering solves ~70% of LLM adaptation tasks for a domain. The remaining 30% require fine-tuning. Three indicators: the model ignores a specific output format even with detailed prompting; the task requires deep knowledge of specialized vocabulary (medicine, law); you need to significantly reduce token costs by replacing a large model with a smaller specialized one.

LoRA and QLoRA are the standard for SFT. LoRA adds trainable low-rank matrices to attention layers. A typical configuration for Llama-3 8B: r=64, lora_alpha=128, target_modules=["q_proj","v_proj","k_proj","o_proj"] yields ~0.8% trainable parameters, training on one A100 40GB. QLoRA adds 4-bit quantization (NF4) and allows fine-tuning 70B models on two A100 40GB, though speed drops by half compared to bf16.

DPO instead of RLHF. Direct Preference Optimization requires only (chosen, rejected) pairs, not scalar reward signals. DPOTrainer from the trl library (Hugging Face) implements it in a few dozen lines.

Common mistake. A dataset of 500 examples, 5 epochs, validation loss 0.8 — seems fine. But on test, the model degrades on general instructions. Cause: catastrophic forgetting. Solution: add 10–20% general instruction-following examples (Alpaca, FLAN) to the training set to preserve original capabilities.

How to Choose a Base Model: 8B or 70B?

For most tasks, fine-tuning an 8B model is sufficient. 70B is needed when deep reasoning is required or the 8B baseline does not reach the required quality even after fine-tuning. Inference cost for Llama-3 8B via vLLM on A100 is efficient; the exact cost depends on volume.

What Does PagedAttention Bring to Production?

vLLM is the first choice for serving open-source models. PagedAttention is the key technical innovation: KV-cache is managed like virtual memory in an OS, without fragmentation. This yields 2–4x higher throughput compared to naive HuggingFace Transformers inference. The vLLM documentation confirms that continuous batching and PagedAttention are the standard for high-load LLM services.

Typical numbers on A100 80GB for Llama-3 8B (bf16): 400–600 req/s, P50 latency 200–400ms, P99 latency 600–900ms at concurrency 64. For 70B on two A100 with tensor parallelism: 80–120 req/s, P99 latency 1.5–2.5s. AWQ or GPTQ quantization reduces memory consumption by 2x with quality loss within 1–3%.

Multi-Agent Systems

Agents are LLMs with access to tools: search, code execution, API calls, database interaction. Common patterns:

• ReAct (Reason + Act): the model reasons → chooses a tool → observes the result → reasons again. LangChain and LlamaIndex implement it out of the box.
• Multi-agent orchestration: multiple specialized agents with a coordinator on top. Example: coordinator → researcher (search + summarization) → coder (code generation and execution) → critic (verification). Tools: AutoGen (Microsoft), CrewAI, custom implementation on LangGraph.

In production, agent systems are non-deterministic. Essential: guardrails, step limits, logging of each step, human-in-the-loop for critical actions.

How We Work: Stages, Timeline, Deliverables

What Is Included

We deliver:

• Technical documentation (model card, configs, deployment instructions)
• Access to infrastructure (code repository, trained weights)
• 1 month of post-deployment support (consultations, bug fixes)
• Customer team training (2–3 sessions on system operation)

Timeline: basic RAG prototype — 1–2 weeks. Fine-tuning with customer data — 3–6 weeks (including data preparation). Production system with monitoring and retraining — 2–4 months. Cost is calculated individually based on data volume, model complexity, and infrastructure requirements.

We guarantee the quality of the final model with performance benchmarks and ongoing monitoring. Our engineers have hands‑on experience with dozens of production LLM systems.

Want to evaluate your project? Leave a request — we will prepare a preliminary summary within 1–2 business days. Or get a consultation on choosing the approach: RAG, fine-tuning, or hybrid — we will tell you what works best for you. Contact us to discuss your LLM development needs. Schedule a free consultation today.