diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 16f28ece2..83ee0a88c 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -31,29 +31,29 @@ jobs:
     steps:
       - uses: actions/checkout@v3
 
-      - name: Cache haskell-stack
-        uses: actions/cache@v3.2.5
+      - uses: actions/cache@v3
+        name: Cache ~/.stack
+        id: cache
         with:
-          path: |
-            ~/.stack
-            .stack-work
-            /home/runner/.local/bin
-          key: ${{ runner.os }}-haskell-stack-${{ hashFiles('**/stack.yaml', '**/package.yaml') }}
+          path: ~/.stack
+          key: ${{ runner.os }}-stack-global-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}
+          restore-keys: |
+            ${{ runner.os }}-stack-global-
+          lookup-only: true
 
       - name: Install haskell-stack
-        uses: haskell/actions/setup@v2.4.3
+        if: steps.cache.outputs.cache-hit != 'true'
+        uses: haskell-actions/setup@v2
         with:
           enable-stack: true
-          stack-no-global: true
           stack-version: "latest"
+          stack-no-global: true
 
-      - name: Build nitta backend dependencies and doctest
+      - name: Install nitta and dev dependencies
+        if: steps.cache.outputs.cache-hit != 'true'
         run: |
           stack build --haddock --test --only-dependencies
-          stack install doctest
-
-      - name: Install weeder
-        run: stack install weeder-2.4.1
+          stack install doctest weeder
 
   nitta:
     runs-on: ubuntu-latest
@@ -62,69 +62,76 @@ jobs:
     steps:
       - uses: actions/checkout@v3
 
-      - name: Cache haskell-stack
-        uses: actions/cache@v3.2.5
+        ############################################################
+        ## Caches
+
+      - uses: actions/cache@v3
+        name: Cache ~/.stack
+        with:
+          path: ~/.stack
+          key: ${{ runner.os }}-stack-global-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}
+          restore-keys: |
+            ${{ runner.os }}-stack-global-
+
+      - uses: actions/cache@v3
+        name: Cache .stack-work
         with:
-          path: |
-            ~/.stack
-            .stack-work
-            /home/runner/.local/bin
-          key: ${{ runner.os }}-haskell-stack-${{ hashFiles('**/stack.yaml', '**/package.yaml') }}
+          path: .stack-work
+          key: ${{ runner.os }}-stack-work-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}-${{ hashFiles('**/*.hs') }}
+          restore-keys: |
+            ${{ runner.os }}-stack-work-
+
+      - uses: actions/cache@v3
+        name: Cache nitta exe
+        with:
+          path: ~/.local/bin
+          key: ${{ runner.os }}-nitta-exe-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}-${{ hashFiles('**/*.hs') }}
+          restore-keys: |
+            ${{ runner.os }}-nitta-exe-
+
+      - uses: actions/cache@v3
+        name: Cache nitta-api-gen
+        with:
+          path: ./web/src/services/gen
+          key: ${{ runner.os }}-nitta-api-gen-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}-${{ hashFiles('**/*.hs') }}
+          restore-keys: |
+            ${{ runner.os }}-nitta-api-gen-
+
+        ############################################################
+        ## Setup dev tools
 
       - name: Install haskell-stack
-        uses: haskell/actions/setup@v2.3.3
+        uses: haskell-actions/setup@v2
         with:
           enable-stack: true
-          stack-no-global: true
           stack-version: "latest"
+          stack-no-global: true
+
+      - name: "Set up HLint"
+        uses: haskell-actions/hlint-setup@v2.4.8
+        with:
+          version: "3.6.1"
+
       - name: Install Icarus Verilog
         run: sudo apt-get install iverilog
 
-      - name: Build & test nitta backend
-        run: stack build --haddock --no-haddock-deps --test --keep-going --test-suite-timeout 600 --copy-bins --coverage
+        ############################################################
+        ## nitta checks
+
+      - name: Build & test nitta
+        run: make build test
 
-      - name: Generate test coverage report
-        run: stack hpc report nitta
+      - name: Check nitta lint and format
+        run: make lint
 
-      - name: Check examples by doctest
-        run: find src -name '*.hs' -exec grep -l '>>>' {} \; | xargs -t -L 1 -P 4 stack exec doctest
+      - name: Make production build
+        run: make build-prod
 
       - name: Generate backend API
         run: stack exec nitta-api-gen -- -v
 
-      - name: Cache node_modules
-        uses: actions/cache@v3.2.5
-        with:
-          path: "**/node_modules"
-          key: ${{ runner.os }}-modules-${{ hashFiles('**/yarn.lock') }}
-
-      - name: Build nitta frontend dependencies
-        working-directory: ./web
-        run: yarn install
-
-      - name: Build frontend
-        working-directory: ./web
-        timeout-minutes: 5
-        run: yarn build
-
-      - name: Run weeder
-        run: weeder
-
-      - uses: abatilo/actions-poetry@v2
-        with:
-          poetry-version: 1.6.1
-      - uses: actions/setup-python@v4
-        with:
-          python-version: 3.11
-          cache: poetry
-      - name: Install Python dependencies
-        working-directory: ./ml/synthesis
-        run: poetry install --no-root
-      - name: Run Python tests
-        working-directory: ./ml/synthesis
-        run: |
-          export PYTHONPATH=$(pwd)/src:$PYTHONPATH
-          poetry run pytest
+        ############################################################
+        ## nitta publish docs & reports
 
       - name: Copy doc to GH_PAGES_DIR
         run: |
@@ -132,7 +139,7 @@ jobs:
           cp -r $(stack path --dist-dir)/doc/html/nitta "$_"
 
       - name: Copy test coverage to GH_PAGES_DIR
-        run: cp -r $(stack path --local-hpc-root)/combined/custom ${{ env.GH_PAGES_DIR }}/hpc
+        run: cp -r $(stack path --local-hpc-root)/combined/all ${{ env.GH_PAGES_DIR }}/hpc
 
       - name: Copy API doc to GH_PAGES_DIR
         run: |
@@ -149,6 +156,100 @@ jobs:
           folder: ${{ env.GH_PAGES_DIR }}
           branch: gh-pages
 
+  nitta-frontend:
+    runs-on: ubuntu-latest
+    needs: nitta
+    timeout-minutes: 15
+    steps:
+      - uses: actions/checkout@v3
+
+        ############################################################
+        ## Caches
+
+      - name: Get yarn cache directory path
+        working-directory: ./web
+        id: yarn-cache-dir-path
+        run: echo "dir=$(yarn cache dir)" >> $GITHUB_OUTPUT
+
+      - uses: actions/cache@v3
+        id: yarn-cache # use this to check for `cache-hit` (`steps.yarn-cache.outputs.cache-hit != 'true'`)
+        with:
+          path: ${{ steps.yarn-cache-dir-path.outputs.dir }}
+          key: ${{ runner.os }}-yarn-${{ hashFiles('**/yarn.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-yarn-
+
+      - uses: actions/cache@v3
+        name: Cache nitta-api-gen
+        with:
+          path: ./web/src/services/gen
+          key: ${{ runner.os }}-nitta-api-gen-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}-${{ hashFiles('**/*.hs') }}
+          restore-keys: |
+            ${{ runner.os }}-nitta-api-gen-
+
+        ############################################################
+        ## nitta ui checks
+
+      - name: Build nitta frontend
+        timeout-minutes: 15
+        run: |
+          yarn --cwd web install
+          make build-ui
+
+  nitta-ml:
+    runs-on: ubuntu-latest
+    needs: nitta
+    timeout-minutes: 15
+    steps:
+      - uses: actions/checkout@v3
+
+        ############################################################
+        ## Caches
+
+      - uses: actions/cache@v3
+        name: Cache nitta exe
+        with:
+          path: ~/.local/bin
+          key: ${{ runner.os }}-nitta-exe-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}-${{ hashFiles('**/*.hs') }}
+          restore-keys: |
+            ${{ runner.os }}-nitta-exe-
+
+      - uses: actions/cache@v3
+        name: Cache nitta-api-gen
+        with:
+          path: ./web/src/services/gen
+          key: ${{ runner.os }}-nitta-api-gen-${{ hashFiles('stack.yaml') }}-${{ hashFiles('package.yaml') }}-${{ hashFiles('**/*.hs') }}
+          restore-keys: |
+            ${{ runner.os }}-nitta-api-gen-
+
+        ############################################################
+        ## Setup dev tools
+
+      - uses: abatilo/actions-poetry@v2
+        with:
+          poetry-version: 1.6.1
+
+      - uses: actions/setup-python@v4
+        with:
+          python-version: 3.11
+          cache: poetry
+
+      - name: Install Python dependencies
+        working-directory: ./ml/synthesis
+        run: poetry install --no-root
+
+        ############################################################
+        ## nitta ml checks
+
+      - name: Run Python tests
+        working-directory: ./ml/synthesis
+        env:
+          NITTA_RUN_COMMAND: nitta
+        run: |
+          export PYTHONPATH=$(pwd)/src:$PYTHONPATH
+
+          poetry run pytest
+
   verilog-formatting:
     timeout-minutes: 5
     runs-on: ubuntu-latest
@@ -163,31 +264,15 @@ jobs:
           cat make_hdl.log
           test "$(grep -ci error make_hdl.log)" -eq 0
 
-  haskell-lint:
-    timeout-minutes: 5
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v3
-
-      - name: "Set up HLint"
-        uses: haskell/actions/hlint-setup@v2.3.3
-        with:
-          version: "3.5"
-
-      - name: "Run HLint"
-        uses: haskell/actions/hlint-run@v2.3.3
-        with:
-          path: .
-          fail-on: suggestion
-
   haskell-formatting:
     timeout-minutes: 5
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v3
 
-      - name: Check formatting
-        uses: fourmolu/fourmolu-action@v8 # fourmolu-0.12.0.0
+      - uses: haskell-actions/run-fourmolu@v9
+        with:
+          version: "0.14.1.0"
 
   typescript-formatting:
     timeout-minutes: 5
diff --git a/Makefile b/Makefile
new file mode 100644
index 000000000..57338afc1
--- /dev/null
+++ b/Makefile
@@ -0,0 +1,93 @@
+HS_SRC_DIR = app src test
+HS_O = --fast
+
+POETRYPATH = ml/synthesis
+PYTHONPATH = ml/synthesis/src
+POETRY = poetry -C $(POETRYPATH)
+PYTHON = PYTHONPATH=$(PYTHONPATH) $(POETRY) run python3
+ML_MODEL_PATH = ml/synthesis/models
+ML_MODEL = $(shell ls ml/synthesis/models -t | grep model | head -n 1)
+
+.PHONY: all build run test format clean
+
+all: format build test lint
+
+############################################################
+## nitta
+############################################################
+
+build:
+	stack build --fast --keep-going \
+		--haddock --no-haddock-deps \
+		--copy-bins --coverage \
+		--test --no-run-tests
+
+build-prod:
+	stack build --ghc-options="-O2" nitta:nitta
+
+test:
+	stack build --coverage --test $(HS_O)
+	find src -name '*.hs' -exec grep -l '>>>' {} \; | xargs -t -L 1 -P 4 stack exec doctest
+
+format:
+	fourmolu -m inplace $(HS_SRC_DIR)
+
+format-check:
+	fourmolu -m check $(HS_SRC_DIR)
+
+lint:
+	hlint $(HS_SRC_DIR)
+	stack exec weeder
+
+clean:
+	stack clean
+
+benchmark:
+	$(PYTHON) $(PYTHONPATH)/scripts/evaluate_nitta_synthesis.py \
+		ml/synthesis/src/scripts/evaluation_configs/demo.json
+
+benchmark-report:
+	$(PYTHON) $(PYTHONPATH)/scripts/compare_evaluations.py \
+		evaluation/*.csv
+
+############################################################
+## nitta ui
+############################################################
+
+ui-build:
+	yarn --cwd web run build
+
+ui-run:
+	yarn --cwd web start
+
+ui-format:
+	yarn --cwd web exec -s prettier -- --write 'web/src/**/*.{ts,tsx}' --ignore-path web/.gitignore
+
+ui-format-check:
+	yarn --cwd web exec -s prettier -- --check 'web/src/**/*.{ts,tsx}' --ignore-path web/.gitignore
+
+############################################################
+## nitta ml
+############################################################
+
+ml-crawl-data:
+	$(PYTHON) $(PYTHONPATH)/scripts/crawl_data_by_tree_sampling_many.py
+
+ml-train-model:
+	$(PYTHON) $(PYTHONPATH)/scripts/train_model.py
+
+ml-format:
+	$(POETRY) run black $(PYTHONPATH)
+
+ml-format-check:
+	$(POETRY) run black --check --diff $(PYTHONPATH)
+
+ml-lint:
+	$(POETRY) run ruff $(PYTHONPATH)
+	$(POETRY) run mypy --config-file ml/synthesis/pyproject.toml $(PYTHONPATH)
+	cd $(POETRYPATH) && poetry run vulture
+
+ml-nitta:
+	echo $(ML_MODEL)
+	$(POETRY) shell
+	MODELS_DIR=$(ML_MODEL_PATH) PYTHONPATH=$(PYTHONPATH) stack exec nitta -- examples/teacup.lua -s $(ML_MODEL) -p=8080 -d=1.2
diff --git a/ml/synthesis/src/components/data_crawling/nitta/nitta_running.py b/ml/synthesis/src/components/data_crawling/nitta/nitta_running.py
index ed15a56de..3437121ee 100644
--- a/ml/synthesis/src/components/data_crawling/nitta/nitta_running.py
+++ b/ml/synthesis/src/components/data_crawling/nitta/nitta_running.py
@@ -144,6 +144,10 @@ async def run_nitta_server(
 ) -> AsyncGenerator[NittaRunResult, None]:
     port = 0  # NITTA will choose a random free port and print it to stdout, we'll parse
 
+    if new_nitta_cmd := os.environ.get("NITTA_RUN_COMMAND"):
+        logger.warn(f"Using NITTA_RUN_COMMAND env var: {new_nitta_cmd}")
+        nitta_run_command = new_nitta_cmd
+
     final_kwargs: dict = dict(
         full_shell_cmd=f"{nitta_run_command} -p={port} {nitta_args} {example}",
     )
diff --git a/ml/synthesis/src/scripts/evaluate_nitta_synthesis.py b/ml/synthesis/src/scripts/evaluate_nitta_synthesis.py
index 73fa1c177..5f5002535 100644
--- a/ml/synthesis/src/scripts/evaluate_nitta_synthesis.py
+++ b/ml/synthesis/src/scripts/evaluate_nitta_synthesis.py
@@ -280,7 +280,7 @@ def _aggregate_and_save_results(results: list[dict], config: EvaluationConfig):
     if not results:
         return
 
-    config.output_dir.mkdir(parents=True, exist_ok=True)
+    Path.mkdir(config.output_dir, parents=True, exist_ok=True)
 
     # dumping raw results just in case something goes wrong while aggregating them
     pickle_output = Path(config.output_dir) / "latest_results.pickle"
diff --git a/ml/synthesis/src/scripts/evaluation_configs/demo.json b/ml/synthesis/src/scripts/evaluation_configs/demo.json
new file mode 100644
index 000000000..4df7a2906
--- /dev/null
+++ b/ml/synthesis/src/scripts/evaluation_configs/demo.json
@@ -0,0 +1,19 @@
+{
+  "output_dir": "evaluation",
+  "nitta_run_command": "stack exec nitta --",
+  "nitta_run_timeout_s": 60,
+  "measurement_tries": 3,
+  "examples": [
+    "fibonacci.lua",
+    "pid.lua",
+    "teacup.lua"
+  ],
+  "constant_args": "-e",
+  "with_ml_backend": true,
+  "evaluated_args": {
+    "preset": {
+      "default-sota": "--method=StateOfTheArt",
+      "default-td": "--method=TopDownByScore --depth-base=1.4"
+    }
+  }
+}