diff --git a/expander_compiler/tests/example_rsa.rs b/expander_compiler/tests/example_rsa.rs
new file mode 100644
index 0000000..334ee02
--- /dev/null
+++ b/expander_compiler/tests/example_rsa.rs
@@ -0,0 +1,201 @@
+use ark_std::test_rng;
+use rand::Rng;
+
+const N_LIMBS: usize = 18;
+const MASK120: u128 = (1 << 120) - 1;
+const MASK60: u128 = (1 << 60) - 1;
+const MASK8: u128 = (1 << 8) - 1;
+const HEX_PER_LIMB: usize = 30;
+
+#[derive(Debug, Clone, Copy, PartialEq)]
+pub struct RSAFieldElement {
+    // an RSA field element is a 2048 bits integer
+    // it is represented as an array of 18 u120 elements, stored each in a u128
+    data: [u128; N_LIMBS],
+}
+
+#[inline]
+fn add_u120_with_carry(a: &u128, b: &u128, carry: &u128) -> (u128, u128) {
+    // a, b, carry are all 120 bits integers, so we can simply add them
+    let mut sum = *a + *b + *carry;
+
+    let carry = sum >> 120;
+    sum = sum & MASK120;
+
+    (sum, carry)
+}
+
+#[inline]
+fn mul_u120_with_carry(a: &u128, b: &u128, carry: &u128) -> (u128, u128) {
+    let a_lo = a & MASK60;
+    let a_hi = a >> 60;
+    let b_lo = b & MASK60;
+    let b_hi = b >> 60;
+    let c_lo = *carry & MASK60;
+    let c_hi = *carry >> 60;
+
+    let tmp_0 = &a_lo * &b_lo + &c_lo;
+    let tmp_1 = &a_lo * &b_hi + &a_hi * &b_lo + c_hi;
+    let tmp_2 = &a_hi * &b_hi;
+
+    let tmp_1_lo = tmp_1 & MASK60;
+    let tmp_1_hi = tmp_1 >> 60;
+
+    let (res, mut c) = add_u120_with_carry(&tmp_0, &(tmp_1_lo << 60), &0u128);
+    c += tmp_1_hi + tmp_2;
+
+    (res, c)
+}
+
+impl RSAFieldElement {
+    pub fn new(data: [u128; N_LIMBS]) -> Self {
+        Self { data }
+    }
+
+    pub fn random(rng: &mut impl Rng) -> Self {
+        let mut data = [0; N_LIMBS];
+        rng.fill(&mut data);
+        data.iter_mut()
+            .take(N_LIMBS - 1)
+            .for_each(|x| *x &= MASK120);
+        data[N_LIMBS - 1] &= MASK8;
+        Self { data }
+    }
+
+    pub fn to_string(&self) -> String {
+        let mut s = String::new();
+        for i in 0..N_LIMBS {
+            s = (&format!("{:030x}", self.data[i])).to_string() + &s;
+        }
+        s
+    }
+
+    pub fn from_string(s: &str) -> Self {
+        let mut data = [0; N_LIMBS];
+        for i in 0..N_LIMBS {
+            data[N_LIMBS - 1 - i] =
+                u128::from_str_radix(&s[i * HEX_PER_LIMB..(i + 1) * HEX_PER_LIMB], 16).unwrap();
+        }
+        Self { data }
+    }
+
+    // assert a + b = result + r * carry
+    pub fn assert_addition(a: &Self, b: &Self, modulus: &Self, carry: &bool, result: &Self) {
+        let mut left_result = [0u128; N_LIMBS]; // for a + b
+        let mut right_result = result.data.clone(); // for result + r * carry
+
+        // First compute a + b
+        let mut c = 0u128;
+        for i in 0..N_LIMBS {
+            let (sum, new_carry) = add_u120_with_carry(&a.data[i], &b.data[i], &c);
+            left_result[i] = sum;
+            c = new_carry;
+        }
+
+        // If carry is true, add modulus to right_result
+        if *carry {
+            let mut c = 0u128;
+            for i in 0..N_LIMBS {
+                let (sum, new_carry) = add_u120_with_carry(&right_result[i], &modulus.data[i], &c);
+                right_result[i] = sum;
+                c = new_carry;
+            }
+        }
+
+        // Assert equality
+        assert!(
+            left_result == right_result,
+            "Addition assertion failed\n{:?}\n{:?}",
+            left_result,
+            right_result
+        );
+    }
+}
+
+#[test]
+fn test_rsa_field_serial() {
+    let mut rng = test_rng();
+    let a = RSAFieldElement::random(&mut rng);
+    let a_str = a.to_string();
+    println!("{:?}", a_str);
+
+    let a2 = RSAFieldElement::from_string(&a_str);
+    assert_eq!(a, a2);
+
+    for _ in 0..100 {
+        let a = RSAFieldElement::random(&mut rng);
+        let a_str = a.to_string();
+        let a2 = RSAFieldElement::from_string(&a_str);
+        assert_eq!(a, a2);
+    }
+}
+
+#[test]
+fn test_u120_add() {
+    let a = MASK120;
+    let b = 1;
+    let carry = 0;
+    let (sum, carry) = add_u120_with_carry(&a, &b, &carry);
+
+    assert_eq!(sum, 0);
+    assert_eq!(carry, 1);
+}
+
+#[test]
+fn test_u120_mul() {
+    let a = MASK120;
+    let b = 8;
+    let carry = 0;
+    let (sum, carry) = mul_u120_with_carry(&a, &b, &carry);
+
+    assert_eq!(sum, 0xfffffffffffffffffffffffffffff8);
+    assert_eq!(carry, 7);
+
+    let a = MASK120;
+    let b = MASK120 - 1;
+    let carry = a;
+    let (sum, carry) = mul_u120_with_carry(&a, &b, &carry);
+
+    assert_eq!(sum, 1);
+    assert_eq!(carry, 0xfffffffffffffffffffffffffffffe);
+}
+
+#[test]
+fn test_assert_rsa_addition() {
+    let mut r = RSAFieldElement::new([MASK120; N_LIMBS]);
+    r.data[N_LIMBS - 1] = MASK8;
+
+    {
+        let a = RSAFieldElement::new([1u128; N_LIMBS]);
+        let b = RSAFieldElement::new([2u128; N_LIMBS]);
+        let result = RSAFieldElement::new([3u128; N_LIMBS]);
+        RSAFieldElement::assert_addition(&a, &b, &r, &false, &result);
+        println!("case 1 passed");
+    }
+
+    {
+        let mut a = RSAFieldElement::new([MASK120 - 1; N_LIMBS]);
+        a.data[N_LIMBS - 1] = MASK8 - 1;
+        let b = RSAFieldElement::new([1u128; N_LIMBS]);
+        let result = RSAFieldElement::new([0u128; N_LIMBS]);
+        println!("a: {:?}", a.to_string());
+        println!("b: {:?}", b.to_string());
+        println!("r: {:?}", r.to_string());
+        println!("result: {:?}", result.to_string());
+        RSAFieldElement::assert_addition(&a, &b, &r, &true, &result);
+        println!("case 2 passed");
+    }
+
+    {
+        let mut a = RSAFieldElement::new([MASK120 - 1; N_LIMBS]);
+        a.data[N_LIMBS - 1] = MASK8 - 1;
+        let b = RSAFieldElement::new([2u128; N_LIMBS]);
+        let result = RSAFieldElement::from_string("000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001000000000000000000000000000001");
+        println!("a: {:?}", a.to_string());
+        println!("b: {:?}", b.to_string());
+        println!("r: {:?}", r.to_string());
+        println!("result: {:?}", result.to_string());
+        RSAFieldElement::assert_addition(&a, &b, &r, &true, &result);
+        println!("case 3 passed");
+    }
+}
diff --git a/expander_compiler/tests/rsa_mul.py b/expander_compiler/tests/rsa_mul.py
new file mode 100644
index 0000000..e8c6c50
--- /dev/null
+++ b/expander_compiler/tests/rsa_mul.py
@@ -0,0 +1,162 @@
+def multiply_mod_254bit_constraint(a_chunks, b_chunks, r_chunks, chunk_size=120):
+    """
+    Compute (a * b) mod r where:
+    - a, b, r are stored as arrays of 17 120-bit chunks
+    - All intermediate calculations must stay under 254 bits
+    - Result should be in the same format (17 120-bit chunks)
+    
+    Parameters:
+    a_chunks: List[int] - 17 elements, each ≤ 2^120
+    b_chunks: List[int] - 17 elements, each ≤ 2^120
+    r_chunks: List[int] - 17 elements, each ≤ 2^120
+    """
+    n = len(a_chunks)  # Should be 17
+    assert len(a_chunks) == len(b_chunks) == len(r_chunks) == 17
+    
+    # Maximum value in each chunk
+    MAX_CHUNK = (1 << chunk_size)
+    
+    def single_chunk_mult(a_i, b_j):
+        """
+        Multiply two 120-bit chunks, ensuring result stays under 254 bits
+        Returns (high, low) tuple where high is the overflow into next chunk
+        """
+        product = a_i * b_j
+        low = product & (MAX_CHUNK - 1)
+        high = product >> chunk_size
+        return high, low
+    
+    # Initialize result array (need extra space for overflow)
+    result = [0] * (2 * n)
+    
+    # Perform multiplication with careful overflow handling
+    for i in range(n):
+        for j in range(n):
+            # Multiply individual chunks
+            high, low = single_chunk_mult(a_chunks[i], b_chunks[j])
+            
+            # Position in result
+            pos = i + j
+            
+            # Add low part
+            result[pos] += low
+            
+            # Handle overflow from low part
+            if result[pos] >= MAX_CHUNK:
+                result[pos + 1] += result[pos] >> chunk_size
+                result[pos] &= (MAX_CHUNK - 1)
+            
+            # Add high part to next chunk
+            result[pos + 1] += high
+            
+            # Handle overflow from high part
+            if result[pos + 1] >= MAX_CHUNK:
+                result[pos + 2] += result[pos + 1] >> chunk_size
+                result[pos + 1] &= (MAX_CHUNK - 1)
+    
+    # Now we need to reduce modulo r
+    def reduce_mod_r():
+        """
+        Reduce the result modulo r, maintaining chunk structure
+        This is a simplified Barrett reduction adapted for our chunk structure
+        """
+        # Convert chunks to a more manageable form for reduction
+        temp_result = result.copy()
+        
+        while any(temp_result[n:]):  # While there are any nonzero high chunks
+            # Find the highest nonzero chunk
+            highest_chunk = 2 * n - 1
+            while highest_chunk >= n and temp_result[highest_chunk] == 0:
+                highest_chunk -= 1
+            
+            if highest_chunk < n:
+                break
+                
+            # Perform the reduction
+            shift = highest_chunk - n + 1
+            for i in range(n):
+                if r_chunks[i] != 0:
+                    for j in range(n):
+                        if i + j + shift < 2 * n:
+                            temp_result[i + j + shift] -= (
+                                (temp_result[highest_chunk] * r_chunks[i]) 
+                                >> (chunk_size * (n - j - 1))
+                            ) & (MAX_CHUNK - 1)
+                            
+            # Normalize chunks
+            for i in range(2 * n - 1):
+                while temp_result[i] < 0:
+                    temp_result[i] += MAX_CHUNK
+                    temp_result[i + 1] -= 1
+                while temp_result[i] >= MAX_CHUNK:
+                    temp_result[i] -= MAX_CHUNK
+                    temp_result[i + 1] += 1
+        
+        return temp_result[:n]
+    
+    return reduce_mod_r()
+
+def test_complex_case():
+    # Create a more complex test case with larger numbers
+    
+    # Helper function to create chunks from a large number
+    def number_to_chunks(num, chunk_size=120, num_chunks=17):
+        chunks = []
+        mask = (1 << chunk_size) - 1
+        for _ in range(num_chunks):
+            chunks.append(num & mask)
+            num >>= chunk_size
+        return chunks
+
+    # Test with some large prime-like numbers
+    # Using numbers that will exercise multiple chunks
+    
+    # Create a large number for testing (about 1000 bits set)
+    a = (1 << 1000) - (1 << 500) + (1 << 200) - (1 << 100) + 12345
+    b = (1 << 900) - (1 << 400) + (1 << 300) - (1 << 50) + 67890
+    r = (1 << 1024) - (1 << 512) + (1 << 256) - (1 << 128) + 11111
+    
+    # Convert to chunks
+    a_chunks = number_to_chunks(a)
+    b_chunks = number_to_chunks(b)
+    r_chunks = number_to_chunks(r)
+    
+    # Perform the modular multiplication
+    result_chunks = multiply_mod_254bit_constraint(a_chunks, b_chunks, r_chunks)
+    
+    # Convert result back to number for verification
+    def chunks_to_number(chunks, chunk_size=120):
+        result = 0
+        for i, chunk in enumerate(chunks):
+            result += chunk << (i * chunk_size)
+        return result
+    
+    # Calculate expected result using regular Python arithmetic
+    expected = (a * b) % r
+    actual = chunks_to_number(result_chunks)
+    
+    print("Test with large numbers:")
+    print(f"a (bits): {a.bit_length()}")
+    print(f"b (bits): {b.bit_length()}")
+    print(f"r (bits): {r.bit_length()}")
+    print("a:", a)
+    print("b:", b)
+    print("r:", r)
+    print("expected:", expected)
+    print("actual:", actual)
+    print("\nFirst few chunks of a:", a_chunks[:3])
+    print("First few chunks of b:", b_chunks[:3])
+    print("First few chunks of r:", r_chunks[:3])
+    print("\nFirst few chunks of result:", result_chunks[:3])
+    print(f"\nExpected result (bits): {expected.bit_length()}")
+    print(f"Actual result (bits): {actual.bit_length()}")
+    print(f"Results match: {expected == actual}")
+    
+    # Additional verification
+    print("\nVerification that no chunk exceeds 120 bits:")
+    max_chunk_size = max(chunk.bit_length() for chunk in result_chunks)
+    print(f"Maximum chunk size in result: {max_chunk_size} bits")
+    print(f"All chunks within 120-bit limit: {max_chunk_size <= 120}")
+
+# Run the test
+test_complex_case()
\ No newline at end of file