diff --git a/puzzles/training/add_space_to_make_sentence.cpp b/puzzles/training/add_space_to_make_sentence.cpp
new file mode 100644
index 0000000..d37b1f9
--- /dev/null
+++ b/puzzles/training/add_space_to_make_sentence.cpp
@@ -0,0 +1,136 @@
+/*
+VIM: let b:cf5build="clang++ -std=c++20 -O2 -pthread -I. {SRC} -o {OUT}"
+VIM: let b:cf5run="{OUT}"
+*/
+#include <iostream>
+#include <exception>
+#include <chrono>
+#include <string>
+#include <string_view>
+#include <unordered_set>
+#include <vector>
+
+/*
+## Problem Description
+
+  Given an input of a dictionary of words and an input string that does not
+  contain spaces, write a method that returns a string with spaces inserted
+  accordingly.
+
+## Expected Completion Time
+
+  For unoptimized recursive solution: 25 minutes
+
+## Problem Complexity Level
+
+  Coding Fluency: Medium
+
+  Problem Solving: Medium
+
+## Questions the candidate might ask
+
+  1) Do you need to find all combinations of sentences?  Always return
+    the sentence with the shorter valid matches.
+  2) What if there are spaces already in the string?  Assume there aren't
+    already spaces.
+  3) What are the valid set of characters (eg. uppercase, symbols)?  Assume
+    dictionary is always all lowercase and are letters.
+  4) What happens if nothing matches in the dictionary?  Return null/none.
+  5) What if there are left over characters in the string?  Return null/none.
+
+## Sample inputs - Expected outputs
+
+  Input: "bloombergisfun", ["bloom","bloomberg","is","fun"]
+  Output: "bloomberg is fun"
+
+## Input corner cases
+
+  Input: "bloombergisfun", ["bloom", "bloomberg","is"]
+  Output: None/Null
+
+  Input: "bloombergisfun", ["bloom", "berg", "bloomberg","is","fun"]
+  Output: "bloom berg is fun"
+
+## Runtime complexity
+
+Recursive Solution: O(2^N)
+This solution O(N*K) where `K` is max length of word in vocab)
+
+Note:
+    hashing of a string is not an O(1) function, but O(N) therefore complexity
+    is O(N*K^2). But it can be made O(N*K).
+
+*/
+
+using dict_t = std::unordered_set<std::string>;
+
+std::string solution(const std::string& str, const dict_t& dict)
+{
+    size_t minw = str.size() +1;
+    size_t maxw = 0;
+    for (const auto& w : dict){
+        minw = std::min(minw, w.size());
+        maxw = std::max(maxw, w.size());
+    }
+
+    std::vector<std::string> memo(str.size()+1, std::string());
+
+    for (size_t i = 0; i < str.size(); ++i){
+        if (i != 0 && memo[i].empty()){
+            continue;
+        }
+        for (size_t j = minw; j <= maxw && i+j <= str.size(); ++j){
+            std::string candidate(str.data()+i, j);
+            auto it = dict.find(candidate);
+            if (it != dict.end()){
+                memo[i+j]=memo[i] + " " + *it;
+            }
+        }
+    }
+
+    return memo[str.size()];
+}
+
+void test(const std::string& str, const dict_t& dict)
+{
+    std::cout << "dict=[";
+    bool first = true;
+    for (const auto& word : dict){
+        if (first){
+            first = false;
+        } else {
+            std::cout <<", ";
+        }
+        std::cout << word;
+    }
+    std::cout << "]\n";
+    std::cout << "str='" << str << "' -> ";
+    std::cout.flush();
+    std::cout << "'" << solution(str, dict) << "'\n\n";
+}
+
+int main ( void )
+{try{
+    auto begin = std::chrono::high_resolution_clock::now();
+
+    test("bloombergisfun", {"bloom","bloomberg","is","fun"});
+    test("bloombergisfun", {"bloom","bloomberg","is"});
+    test("bloombergisfun", {"bloom", "berg", "bloomberg","is","fun"});
+
+    auto end = std::chrono::high_resolution_clock::now();
+    std::chrono::duration<double> seconds = end - begin;
+    std::cout << "Time: " << seconds.count() << std::endl;
+    return 0;
+}
+catch ( const std::exception& e )
+{
+    std::cerr << std::endl
+            << "std::exception(\"" << e.what() << "\")." << std::endl;
+    return 2;
+}
+catch ( ... )
+{
+    std::cerr  << std::endl
+            << "unknown exception." << std::endl;
+    return 1;
+}}