diff --git a/csv2/struct.go b/csv2/struct.go
new file mode 100644
index 0000000..b2bd5fb
--- /dev/null
+++ b/csv2/struct.go
@@ -0,0 +1,154 @@
+package csv2
+
+import (
+	"encoding/csv"
+	"io"
+	"reflect"
+	"strconv"
+
+	"github.com/juju/errors"
+	"github.com/localhots/gobelt/reflect2"
+)
+
+// Reader is a wrapper for standard library CSV Reader that allows unmarshalling
+// into a slice of structs.
+type Reader struct {
+	TagName               string
+	ColumnNamesInFirstRow bool
+	Reader                *csv.Reader
+	cols                  []string
+}
+
+const defaultTagName = "csv"
+
+// NewReader creates a new reader from a standard CSV Reader.
+func NewReader(r *csv.Reader) *Reader {
+	return &Reader{
+		TagName:               defaultTagName,
+		ColumnNamesInFirstRow: true,
+		Reader:                r,
+	}
+}
+
+// SetColumnNames assigns column names. Use this function if column names are
+// not provided in the first row.
+func (r *Reader) SetColumnNames(cols []string) {
+	if r.ColumnNamesInFirstRow {
+		panic("Should not assign column names when they are expected in the first row")
+	}
+	r.cols = cols
+}
+
+// Load reads CSV contents and unmarshals them into given destination.
+// Destination value must be a pointer to a slice of structs.
+func (r *Reader) Load(dest interface{}) error {
+	destT := reflect.TypeOf(dest)
+	if destT.Kind() != reflect.Ptr ||
+		destT.Elem().Kind() != reflect.Slice ||
+		destT.Elem().Elem().Kind() != reflect.Struct {
+		return errors.New("Destination must be a pointer to a slice of structs")
+	}
+
+	if r.cols == nil {
+		if r.ColumnNamesInFirstRow {
+			cols, err := r.Reader.Read()
+			if err != nil {
+				return errors.Annotate(err, "Failed to read column names from first row")
+			}
+			r.cols = cols
+		} else {
+			return errors.New("Column names are not defined")
+		}
+	}
+
+	destV := reflect.ValueOf(dest).Elem()
+	valT := destT.Elem().Elem()
+	colIndex := reflect2.AssociateColumns(valT, r.TagName, r.cols)
+
+	for {
+		row, err := r.Reader.Read()
+		if err != nil {
+			if err == io.EOF {
+				break
+			}
+			return errors.Annotate(err, "Failed to read CSV row")
+		}
+
+		val := reflect.New(valT).Elem()
+		for iCol, iField := range colIndex {
+			err := unmarshal(row[iCol], val.Field(iField))
+			if err != nil {
+				return errors.Annotate(err, "Failed to process CSV row")
+			}
+		}
+
+		destV.Set(reflect.Append(destV, val))
+	}
+	return nil
+}
+
+func unmarshal(v string, dest reflect.Value) error {
+	switch dest.Kind() {
+	case reflect.String:
+		dest.SetString(v)
+	case reflect.Bool:
+		b, err := strconv.ParseBool(v)
+		if err != nil {
+			return unmarshalError(err, v, dest.Kind())
+		}
+		dest.SetBool(b)
+	case reflect.Int, reflect.Int64:
+		return unmarshalInt(v, dest, 64)
+	case reflect.Int8:
+		return unmarshalInt(v, dest, 8)
+	case reflect.Int16:
+		return unmarshalInt(v, dest, 16)
+	case reflect.Int32:
+		return unmarshalInt(v, dest, 32)
+	case reflect.Uint, reflect.Uint64:
+		return unmarshalUint(v, dest, 64)
+	case reflect.Uint8:
+		return unmarshalUint(v, dest, 8)
+	case reflect.Uint16:
+		return unmarshalUint(v, dest, 16)
+	case reflect.Uint32:
+		return unmarshalUint(v, dest, 32)
+	case reflect.Float32:
+		return unmarshalFloat(v, dest, 32)
+	case reflect.Float64:
+		return unmarshalFloat(v, dest, 64)
+	}
+
+	return nil
+}
+
+func unmarshalInt(v string, dest reflect.Value, bitSize int) error {
+	i, err := strconv.ParseInt(v, 10, bitSize)
+	if err != nil {
+		return unmarshalError(err, v, dest.Kind())
+	}
+	dest.SetInt(i)
+	return nil
+}
+
+func unmarshalUint(v string, dest reflect.Value, bitSize int) error {
+	i, err := strconv.ParseUint(v, 10, bitSize)
+	if err != nil {
+		return unmarshalError(err, v, dest.Kind())
+	}
+	dest.SetUint(i)
+	return nil
+}
+
+func unmarshalFloat(v string, dest reflect.Value, bitSize int) error {
+	f, err := strconv.ParseFloat(v, bitSize)
+	if err != nil {
+		return unmarshalError(err, v, dest.Kind())
+	}
+	dest.SetFloat(f)
+	return nil
+}
+
+func unmarshalError(err error, v string, k reflect.Kind) error {
+	return errors.Annotatef(err, "Can't unmarshal %q into value of type %s", v, k)
+}
diff --git a/csv2/struct_test.go b/csv2/struct_test.go
new file mode 100644
index 0000000..cd2dfe8
--- /dev/null
+++ b/csv2/struct_test.go
@@ -0,0 +1,42 @@
+package csv2
+
+import (
+	"encoding/csv"
+	"strings"
+	"testing"
+
+	"github.com/google/go-cmp/cmp"
+)
+
+func TestLoad(t *testing.T) {
+	type specie struct {
+		Name      string  `csv:"name"`
+		FavFood   string  `csv:"fav_food"`
+		Age       uint16  `csv:"age"`
+		Weight    float32 `csv:"weight"`
+		Available bool    `csv:"available"`
+	}
+	body := `name,available,fav_food,weight,age
+Alice,false,Bananas,19.22,5
+Frank,true,Burrito,14,9
+Joel,true,Pesto,32.5,21`
+	exp := []specie{
+		{Name: "Alice", FavFood: "Bananas", Age: 5, Weight: 19.22, Available: false},
+		{Name: "Frank", FavFood: "Burrito", Age: 9, Weight: 14, Available: true},
+		{Name: "Joel", FavFood: "Pesto", Age: 21, Weight: 32.5, Available: true},
+	}
+
+	csvReader := csv.NewReader(strings.NewReader(body))
+	r := NewReader(csvReader)
+	r.ColumnNamesInFirstRow = true
+
+	var out []specie
+	err := r.Load(&out)
+	if err != nil {
+		t.Fatalf("Unexpected error: %v", err)
+	}
+
+	if !cmp.Equal(exp, out) {
+		t.Errorf("Result value is different: %s", cmp.Diff(exp, out))
+	}
+}