ABB Software Engineer and Graduate Engineer campus loops are resume-heavy: an online test filters the pool, then technical panels interrogate projects, graphs/OOP, and domain fit before a behavioural HR close. We reorganized public candidate write-ups (including pooled IIT drives) into round tables, outcome notes, and prep takeaways - treat this as a prep map, not a script.
Structured problem → approach → impact without slide decks
Technical interview experience 1 - Software Engineer, pooled IIT campus
Technical interview experience 1
Candidate profile: EE/CSE pool campus (3rd-gen IITs); Software Engineer role. Funnel reported: ~120-130 sat OA → 26 shortlisted → ~15 → ~10 → 7 offers (4 + 3 across colleges).
Round 1 - Online test
Sec 1: Aptitude and reasoning (~20 min)
Sec 2: Coding (~1 hr) - Java / C++ / Python / C
Reported coding prompts: (1) print path between any two nodes in a binary tree; (2) count how many times a given string is swapped with neighbours while sorting by ASCII sum of first three characters (bubble-sort adjacent swaps); (3) total unit cells covered by all given rectangles
Sec 3: Choose C#/C++/.NET or Java MCQs (~30 min) - OOP and output-style questions
Tip: Candidates who only crushed coding and ignored MCQs/aptitude were not shortlisted in this report
Round 2 - Technical (~30 min)
Open prompt: 5 minutes to talk about anything → projects
Graphs: why graphs matter; BFS vs DFS; MST with a real-life example (e.g. connecting electric poles)
Arduino / electrical project basics when on the resume
Draw a binary tree (Paint) and show DFS and BFS order
Basic OOP including multiple inheritance
Fields of interest inside ABB
Result: Cleared (~15 advanced in this account)
Round 3 - Technical (~30 min, finished technical in ~15)
Socket-programming project: client-server architecture, real-life examples, multithreading for performance
C++ OOP: multiple vs multilevel inheritance; diamond problem
Algorithm discussion (idea + complexity, coding optional): count the first repeating character in a string
Hobbies / closing questions about how OS/networks show up in ABB work
Result: Cleared (~10 advanced)
Round 4 - Surprise DSA + HR blend (~30 min)
Expected soft HR; instead got: given an input string, print all valid anagrams stored in a huge file (~1M words)
Discussion path reported: generate anagrams → hash map; improve via count arrays O(26); then precompute count-array keys for the whole file into a map for O(1)/O(log n) lookup (interviewer hinted “lots of memory”)
Strengths / weaknesses with situations
Round 5 - HR (~20 min)
Introduce yourself
Situational: teammate not responding to email/messages - what do you do?
Conflict between team members; biggest failure; leadership experience
Where do you see yourself in 5 years?
Final result: Offer (among 7 selects in the drive)
Takeaways
What decided it: Honest project depth (PostgreSQL/Angular/sockets) plus clear BFS/DFS/MST and OOP answers under follow-ups
Common fail: Crushing coding while ignoring aptitude/language MCQs, or going silent on the file-scale anagram follow-up
How to practise: Dry-run resume projects aloud; solve path-between-nodes and count-array anagram lookup on paper
Candidate profile: EEE/ECE campus hire targeting GET (not software track). Process shape from public fresher guides: online aptitude + branch technical → optional GD → long technical → substantive HR.
Round 1 - Online test (~110 min reported)
Aptitude, verbal, and analytical sections timed as one block
Branch technical: machines, power systems, control basics, instrumentation MCQs for EE; separate CS track on software drives
Tip: FacePrep and placementpapers.app both note that technical MCQs eliminate more candidates than verbal - do not spend the whole paper on easy aptitude and rush the core section
Round 2 - Technical (45-75 min)
Panel asks candidate to pick 2-3 domains for depth (e.g. power electronics + control systems)
Resume walk: final-year project on induction motor control, internship at a panel builder
Follow-ups reported in prep lists: star-delta starting, AC vs DC preference in distribution, cache memory role/limitations, DBMS vs file storage when IT topics appear
Result: Cleared when candidate re-derived formulas instead of reciting memorised answers
Why ABB over IT services? - panel treats this as a real filter for core-engineering intent
Safety scenario: noticed unsafe practice - how would you respond?
Team-under-pressure story; five-year plan inside engineering, not generic “manager”
Final result: Offer (package candidate-reported in ₹7-13 LPA band - confirm on letter)
Key takeaways: GET loops reward domain depth + safety mindset + relocation honesty. Software-track candidates should study Technical interview experience 1 on this page instead of memorising motor formulas alone.
HR / Management Trainee notes
HR and MT-style extras
Older Management Trainee campus reports add: ~1 hour resume-line technical screen, then a chart-paper business case (markers only, no PPT), a long personal HR (~1.5 hr), and a GD (e.g. greenhouse effect, ~45 min). Soft prompts overlap Software Engineer HR: conflict, failure, leadership, 5-year plan.
Tip: For chart-paper cases, structure problem → options → recommendation → risks - visual polish does not matter.
Sample answers for high-frequency questions
Use as frames - panels probe follow-ups.
Q. Explain BFS vs DFS and when you would use each. A. BFS explores level by level with a queue - shortest path in unweighted graphs, nearest neighbour style search. DFS goes deep with a stack/recursion - cycle detection, topological sort, connected components, path existence. Prefer BFS when distance/layers matter; DFS when you need to explore full branches or use less memory on wide shallow graphs.
Q. What is an MST, with a real example? A. A minimum spanning tree connects all vertices with minimum total edge weight and no cycles. Example from an ABB-style report: laying cable between electric poles so every pole is connected at minimum wire cost. Algorithms: Kruskal (sort edges + Union-Find) or Prim (grow from a start vertex).
Q. First repeating character in a string - approach and complexity. A. Scan left to right with a hash set of seen characters; the first character already in the set is the answer. O(n) time, O(alphabet) space. Variants: first non-repeating uses a frequency map then a second pass.
Q. Find all anagrams of a word in a large dictionary file. A. Build a signature for the query (sorted characters or 26-length count key). Precompute the same signature for every dictionary word into a hash map from signature → list of words. Lookup is average O(1) after O(total dictionary characters) preprocessing. Comparing unsorted permutations without indexing does not scale to ~1M words.
Q. Edge computing vs cloud - why both exist at ABB-style IoT stacks? A. Edge processes data near sensors for low latency and bandwidth savings (control loops, local filtering). Cloud aggregates fleet data for heavy analytics, model training and central dashboards. Production systems usually push summaries or exceptions upstream rather than raw high-frequency streams.
Dense prep checklist
Know every project line: stack choices, multithreading, client-server, what you would change
Graphs: BFS/DFS traversals on paper, MST intuition
OOP: inheritance diamond problem, interfaces vs abstract classes
Domain skim: ABB Ability, IoT, edge vs cloud if the PPT mentions them
HR: one failure STAR, one conflict STAR, clear relocation answer
These links are research inputs we used to map round structure and recurring questions - not copied page mirrors. We reorganized reports into process tables, outcome notes, and prep takeaways for ABB.
