diff --git a/stubgen/inferrer.py b/stubgen/inferrer.py
new file mode 100644
index 0000000..4ab5a9e
--- /dev/null
+++ b/stubgen/inferrer.py
@@ -0,0 +1,533 @@
+"""Type inference engine for Python code analysis."""
+
+import ast
+import re
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Set, Tuple
+
+from stubgen.parser import (
+    FileInfo, Function, Variable, Class, TypeHint
+)
+
+
+class TypeCandidate:
+    """Represents a candidate type with confidence score."""
+    
+    def __init__(
+        self,
+        type_str: str,
+        confidence: float,
+        source: str = "inference"
+    ):
+        self.type_str = type_str
+        self.confidence = confidence
+        self.source = source
+    
+    def __repr__(self):
+        return f"TypeCandidate({self.type_str}, confidence={self.confidence})"
+
+
+class ReturnTypeAnalyzer:
+    """Analyzes return statements to infer function return types."""
+    
+    def __init__(self, func: Function):
+        self.func = func
+    
+    def analyze(self) -> Optional[TypeCandidate]:
+        return_types: Set[str] = set()
+        
+        for stmt in self.func.body:
+            result = self._analyze_statement(stmt)
+            if result:
+                return_types.add(result)
+        
+        if not return_types:
+            if self._has_yield(stmt for stmt in self.func.body):
+                return TypeCandidate("Generator", 0.9, "yield")
+            return None
+        
+        if len(return_types) == 1:
+            return_type = return_types.pop()
+            confidence = 1.0
+            if return_type == "None":
+                confidence = 0.7
+            return TypeCandidate(return_type, confidence, "return")
+        
+        if return_types == {"None", "bool"}:
+            return TypeCandidate("Optional[bool]", 0.8, "return")
+        if return_types == {"None", "int"}:
+            return TypeCandidate("Optional[int]", 0.8, "return")
+        if return_types == {"None", "str"}:
+            return TypeCandidate("Optional[str]", 0.8, "return")
+        if return_types == {"None"}:
+            return TypeCandidate("None", 1.0, "return")
+        
+        return TypeCandidate("Any", 0.5, "return")
+    
+    def _has_yield(self, statements) -> bool:
+        for stmt in statements:
+            if isinstance(stmt, (ast.Yield, ast.YieldFrom)):
+                return True
+            for child in ast.walk(stmt):
+                if isinstance(child, (ast.Yield, ast.YieldFrom)):
+                    return True
+        return False
+    
+    def _analyze_statement(self, stmt: ast.stmt) -> Optional[str]:
+        if isinstance(stmt, ast.Return):
+            if stmt.value is None:
+                return "None"
+            return self._infer_type(stmt.value)
+        elif isinstance(stmt, ast.Expr) and isinstance(stmt.value, ast.Constant):
+            return self._infer_type(stmt.value)
+        elif isinstance(stmt, ast.Assign):
+            if stmt.value:
+                return self._infer_type(stmt.value)
+        elif isinstance(stmt, (ast.If, ast.For, ast.While)):
+            for child in stmt.body:
+                result = self._analyze_statement(child)
+                if result:
+                    return result
+            if hasattr(stmt, 'orelse'):
+                for child in stmt.orelse:
+                    result = self._analyze_statement(child)
+                    if result:
+                        return result
+        elif isinstance(stmt, ast.Try):
+            for child in stmt.body:
+                result = self._analyze_statement(child)
+                if result:
+                    return result
+        return None
+    
+    def _infer_type(self, node: ast.expr) -> str:
+        if isinstance(node, ast.Constant):
+            value = node.value
+            if isinstance(value, bool):
+                return "bool"
+            elif isinstance(value, int):
+                return "int"
+            elif isinstance(value, float):
+                return "float"
+            elif isinstance(value, str):
+                return "str"
+            elif value is None:
+                return "None"
+        elif isinstance(node, ast.List):
+            if node.elts:
+                types = {self._infer_type(elt) for elt in node.elts}
+                if len(types) == 1:
+                    return f"List[{types.pop()}]"
+            return "List[Any]"
+        elif isinstance(node, ast.Tuple):
+            if node.elts:
+                types = [self._infer_type(elt) for elt in node.elts]
+                return f"Tuple[{', '.join(types)}]"
+            return "Tuple[()]"
+        elif isinstance(node, ast.Dict):
+            return "Dict[Any, Any]"
+        elif isinstance(node, ast.Set):
+            if node.elts:
+                types = {self._infer_type(elt) for elt in node.elts}
+                if len(types) == 1:
+                    return f"Set[{types.pop()}]"
+            return "Set[Any]"
+        elif isinstance(node, ast.Name):
+            return self._infer_name_type(node)
+        elif isinstance(node, ast.BinOp):
+            left_type = self._infer_type(node.left)
+            right_type = self._infer_type(node.right)
+            if left_type == right_type:
+                return left_type
+            if left_type in ("int", "float") and right_type in ("int", "float"):
+                return "float" if "float" in (left_type, right_type) else "int"
+            return "Any"
+        elif isinstance(node, ast.Call):
+            if isinstance(node.func, ast.Name):
+                func_name = node.func.id
+                return self._infer_call_type(func_name, node)
+            return "Any"
+        elif isinstance(node, ast.Attribute):
+            return "Any"
+        return "Any"
+    
+    def _infer_name_type(self, node: ast.Name) -> str:
+        if node.id in ("True", "False"):
+            return "bool"
+        return "Any"
+    
+    def _infer_call_type(self, func_name: str, node: ast.Call) -> str:
+        type_hints = {
+            "len": "int",
+            "str": "str",
+            "int": "int",
+            "float": "float",
+            "bool": "bool",
+            "list": "list",
+            "dict": "dict",
+            "set": "set",
+            "tuple": "tuple",
+            "range": "range",
+            "enumerate": "enumerate",
+            "zip": "zip",
+            "map": "map",
+            "filter": "filter",
+            "sorted": "list",
+            "reversed": "list",
+            "type": "type",
+            "isinstance": "bool",
+            "hasattr": "bool",
+            "getattr": "Any",
+            "open": "IO",
+        }
+        return type_hints.get(func_name, "Any")
+
+
+class DefaultValueAnalyzer:
+    """Analyzes default values to infer parameter types."""
+    
+    def analyze(self, default: Optional[ast.expr]) -> Optional[TypeCandidate]:
+        if default is None:
+            return None
+        
+        default_type = self._infer_type(default)
+        
+        confidence_map = {
+            "None": 0.7,
+            "str": 0.8,
+            "int": 0.8,
+            "float": 0.8,
+            "bool": 0.8,
+            "list": 0.6,
+            "dict": 0.6,
+            "set": 0.6,
+            "tuple": 0.6,
+            "Any": 0.5,
+        }
+        
+        confidence = confidence_map.get(default_type, 0.6)
+        return TypeCandidate(default_type, confidence, "default")
+    
+    def _infer_type(self, node: ast.expr) -> str:
+        if isinstance(node, ast.Constant):
+            value = node.value
+            if isinstance(value, bool):
+                return "bool"
+            elif isinstance(value, int):
+                return "int"
+            elif isinstance(value, float):
+                return "float"
+            elif isinstance(value, str):
+                return "str"
+            elif value is None:
+                return "None"
+        elif isinstance(node, ast.ListConstant):
+            return "list"
+        elif isinstance(node, ast.Dict):
+            return "dict"
+        elif isinstance(node, ast.Tuple):
+            return "tuple"
+        elif isinstance(node, ast.Set):
+            return "set"
+        elif isinstance(node, ast.Name):
+            if node.id in ("None", "True", "False"):
+                return "None" if node.id == "None" else "bool"
+        elif isinstance(node, ast.NameConstant):
+            if node.value is None:
+                return "None"
+            elif node.value is bool:
+                return "bool"
+        return "Any"
+
+
+class AssignmentAnalyzer:
+    """Analyzes variable assignments to infer types."""
+    
+    def __init__(self, file_info: FileInfo):
+        self.file_info = file_info
+        self.assignments: Dict[str, str] = {}
+    
+    def analyze(self) -> Dict[str, TypeCandidate]:
+        results = {}
+        
+        for var in self.file_info.variables:
+            if var.value:
+                inferred = self._infer_type(var.value)
+                confidence = 0.8 if inferred != "Any" else 0.5
+                results[var.name] = TypeCandidate(inferred, confidence, "assignment")
+            elif var.annotation:
+                results[var.name] = TypeCandidate(
+                    var.annotation.to_string(),
+                    1.0,
+                    "annotation"
+                )
+        
+        return results
+    
+    def _infer_type(self, node: ast.expr) -> str:
+        if isinstance(node, ast.Constant):
+            value = node.value
+            if isinstance(value, bool):
+                return "bool"
+            elif isinstance(value, int):
+                return "int"
+            elif isinstance(value, float):
+                return "float"
+            elif isinstance(value, str):
+                return "str"
+            elif value is None:
+                return "None"
+        elif isinstance(node, ast.List):
+            if node.elts:
+                types = {self._infer_type(elt) for elt in node.elts}
+                if len(types) == 1:
+                    return f"List[{types.pop()}]"
+            return "list"
+        elif isinstance(node, ast.Tuple):
+            return "tuple"
+        elif isinstance(node, ast.Dict):
+            return "dict"
+        elif isinstance(node, ast.Set):
+            return "set"
+        elif isinstance(node, ast.Call):
+            if isinstance(node.func, ast.Name):
+                func_name = node.func.id
+                type_map = {
+                    "list": "list",
+                    "dict": "dict",
+                    "set": "set",
+                    "tuple": "tuple",
+                    "int": "int",
+                    "str": "str",
+                    "float": "float",
+                    "bool": "bool",
+                }
+                return type_map.get(func_name, "Any")
+            return "Any"
+        return "Any"
+
+
+class DocstringParser:
+    """Parses docstrings for type hints (Google and NumPy style)."""
+    
+    def __init__(self, docstring: str):
+        self.docstring = docstring or ""
+    
+    def parse_params(self) -> Dict[str, str]:
+        params = {}
+        
+        google_match = re.search(
+            r'Args:\s*\n(.*?)(?:\n\n|\n[A-Z]|\Z)',
+            self.docstring,
+            re.DOTALL
+        )
+        if google_match:
+            args_section = google_match.group(1)
+            for match in re.finditer(r'(\w+)\s*\(([^)]+)\)\s*:\s*(.*)', args_section):
+                param_name = match.group(1)
+                param_type = match.group(2).strip()
+                params[param_name] = param_type
+        
+        numpy_match = re.search(
+            r'Parameters\s*\n\s*[-]+\s*\n(.*?)(?:\n\n|\n[A-Z]|\Z)',
+            self.docstring,
+            re.DOTALL
+        )
+        if numpy_match:
+            params_section = numpy_match.group(1)
+            for match in re.finditer(r'(\w+)\s*(?:[-\s]+)?(?:type| dtype):\s*(.*)', params_section):
+                param_name = match.group(1)
+                param_type = match.group(2).strip()
+                if param_name not in params:
+                    params[param_name] = param_type
+        
+        return params
+    
+    def parse_returns(self) -> Optional[str]:
+        google_match = re.search(
+            r'Returns?:\s*\n\s*(?:.*?):\s*(.*?)(?:\n\n|\n[A-Z]|\Z)',
+            self.docstring,
+            re.DOTALL
+        )
+        if google_match:
+            return google_match.group(1).strip()
+        
+        numpy_match = re.search(
+            r'Returns?\s*\n\s*[-]+\s*\n\s*(?:.*?):\s*(.*?)(?:\n\n|\Z)',
+            self.docstring,
+            re.DOTALL
+        )
+        if numpy_match:
+            return numpy_match.group(1).strip()
+        
+        return None
+    
+    def parse_attributes(self) -> Dict[str, str]:
+        attrs = {}
+        
+        google_match = re.search(
+            r'Attributes:\s*\n(.*?)(?:\n\n|\n[A-Z]|\Z)',
+            self.docstring,
+            re.DOTALL
+        )
+        if google_match:
+            attrs_section = google_match.group(1)
+            for match in re.finditer(r'(\w+)\s*\(([^)]+)\)\s*:\s*(.*)', attrs_section):
+                attr_name = match.group(1)
+                attr_type = match.group(2).strip()
+                attrs[attr_name] = attr_type
+        
+        return attrs
+
+
+class Inferrer:
+    """Main type inference engine that combines all analyzers."""
+    
+    def __init__(self, infer_depth: int = 3):
+        self.infer_depth = infer_depth
+    
+    def infer_function_signature(self, func: Function, docstring: str = None) -> Dict[str, Any]:
+        """Infer complete function signature including parameter and return types."""
+        args_info = {}
+        
+        docstring_parser = DocstringParser(docstring) if docstring else DocstringParser("")
+        doc_params = docstring_parser.parse_params()
+        
+        for arg in func.args.args:
+            arg_info = {"name": arg.arg}
+            
+            if arg.annotation:
+                arg_info["type"] = arg.annotation
+                arg_info["source"] = "annotation"
+                arg_info["confidence"] = 1.0
+            elif arg.arg in doc_params:
+                arg_info["type"] = doc_params[arg.arg]
+                arg_info["source"] = "docstring"
+                arg_info["confidence"] = 0.9
+            else:
+                default_idx = len(func.args.args) - len(func.args.defaults) + func.args.args.index(arg)
+                if default_idx >= 0 and default_idx < len(func.args.defaults):
+                    default_analyzer = DefaultValueAnalyzer()
+                    candidate = default_analyzer.analyze(func.args.defaults[default_idx])
+                    if candidate:
+                        arg_info["type"] = candidate.type_str
+                        arg_info["source"] = "default"
+                        arg_info["confidence"] = candidate.confidence
+                    else:
+                        arg_info["type"] = "Any"
+                        arg_info["source"] = "none"
+                        arg_info["confidence"] = 0.5
+                else:
+                    arg_info["type"] = "Any"
+                    arg_info["source"] = "none"
+                    arg_info["confidence"] = 0.5
+            
+            args_info[arg.arg] = arg_info
+        
+        for kwonly in func.args.kwonlyargs:
+            arg_info = {"name": kwonly.arg}
+            if kwonly.annotation:
+                arg_info["type"] = kwonly.annotation
+                arg_info["source"] = "annotation"
+                arg_info["confidence"] = 1.0
+            else:
+                arg_info["type"] = "Any"
+                arg_info["source"] = "none"
+                arg_info["confidence"] = 0.5
+            args_info[kwonly.arg] = arg_info
+        
+        if func.args.vararg:
+            args_info[func.args.vararg.arg] = {
+                "name": func.args.vararg.arg,
+                "type": "*Tuple[Any, ...]",
+                "source": "vararg",
+                "confidence": 1.0
+            }
+        
+        if func.args.kwarg:
+            args_info[func.args.kwarg.arg] = {
+                "name": func.args.kwarg.arg,
+                "type": "**Dict[str, Any]",
+                "source": "kwarg",
+                "confidence": 1.0
+            }
+        
+        return_type_info = {"type": "None", "source": "default", "confidence": 1.0}
+        
+        if func.returns:
+            return_type_info = {
+                "type": func.returns.to_string(),
+                "source": "annotation",
+                "confidence": 1.0
+            }
+        else:
+            return_analyzer = ReturnTypeAnalyzer(func)
+            candidate = return_analyzer.analyze()
+            if candidate:
+                return_type_info = {
+                    "type": candidate.type_str,
+                    "source": candidate.source,
+                    "confidence": candidate.confidence
+                }
+            elif docstring:
+                doc_returns = docstring_parser.parse_returns()
+                if doc_returns:
+                    return_type_info = {
+                        "type": doc_returns,
+                        "source": "docstring",
+                        "confidence": 0.9
+                    }
+        
+        return {
+            "args": args_info,
+            "returns": return_type_info,
+            "is_async": func.is_async,
+            "decorators": func.decorators
+        }
+    
+    def infer_file(self, file_info: FileInfo) -> Dict[str, Any]:
+        """Infer types for all components in a file."""
+        results = {
+            "functions": {},
+            "classes": {},
+            "variables": {}
+        }
+        
+        for func in file_info.functions:
+            results["functions"][func.name] = self.infer_function_signature(
+                func, file_info.docstring
+            )
+        
+        for cls in file_info.classes:
+            class_info = {
+                "name": cls.name,
+                "bases": cls.bases,
+                "methods": {},
+                "attributes": {}
+            }
+            
+            for method in cls.methods:
+                class_info["methods"][method.name] = self.infer_function_signature(
+                    method, None
+                )
+            
+            for attr in cls.attributes:
+                if attr.annotation:
+                    class_info["attributes"][attr.name] = {
+                        "type": attr.annotation.to_string(),
+                        "source": "annotation",
+                        "confidence": 1.0
+                    }
+                else:
+                    class_info["attributes"][attr.name] = {
+                        "type": "Any",
+                        "source": "none",
+                        "confidence": 0.5
+                    }
+            
+            results["classes"][cls.name] = class_info
+        
+        assignment_analyzer = AssignmentAnalyzer(file_info)
+        results["variables"] = assignment_analyzer.analyze()
+        
+        return results