SQL Server 2017

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.

CREATE TABLE test (interview_id INT, interviewee_uid INT, interviewer_uid INT, idate DATE)
INSERT INTO test VALUES
(1, 1, 11, '2019-07-29'),
(2, 1, 12, '2019-07-28'),
(3, 1, 13, '2019-07-27'),
(4, 1, 14, '2019-07-26'),
(5, 1, 12, '2019-07-25'),
(6, 2, 11, '2019-07-29'),
(7, 2, 13, '2019-07-28'),
(8, 3, 12, '2019-07-29'),
(9, 3, 12, '2019-07-28'),
(10, 3, 12, '2019-07-27'),
(11, 3, 13, '2019-07-26'),
(12, 4, 11, '2019-07-29'),
(13, 4, 11, '2019-07-28')
13 rows affected

WITH cte AS (
SELECT *,
       DENSE_RANK() OVER (PARTITION BY interviewee_uid ORDER BY interviewer_uid) iid,
       ROW_NUMBER() OVER (PARTITION BY interviewee_uid, interviewer_uid ORDER BY idate DESC) did
FROM test
)
SELECT * 
FROM cte
WHERE did = 1
  AND iid <= 3 
;


interview_id
interviewee_uid
interviewer_uid
idate
iid
did


1
1
11
29/07/2019 00:00:00
1
1

2
1
12
28/07/2019 00:00:00
2
1

3
1
13
27/07/2019 00:00:00
3
1

6
2
11
29/07/2019 00:00:00
1
1

7
2
13
28/07/2019 00:00:00
2
1

8
3
12
29/07/2019 00:00:00
1
1

11
3
13
26/07/2019 00:00:00
2
1

12
4
11
29/07/2019 00:00:00
1
1

interview_id	interviewee_uid	interviewer_uid	idate	iid	did
1	1	11	29/07/2019 00:00:00	1	1
2	1	12	28/07/2019 00:00:00	2	1
3	1	13	27/07/2019 00:00:00	3	1
6	2	11	29/07/2019 00:00:00	1	1
7	2	13	28/07/2019 00:00:00	2	1
8	3	12	29/07/2019 00:00:00	1	1
11	3	13	26/07/2019 00:00:00	2	1
12	4	11	29/07/2019 00:00:00	1	1