Hari Mahajan — Portfolio

Projects

Certified

Self-directed project

Predictive Maintenance: Industrial ML for Fault Detection

Condition-Monitoring Engineer · AI / ML · August 2026

68/ 100

A ProoV case study · educational project, not employment

Built a compact rolling-bearing fault detector end to end: loaded and interpreted vibration snapshots, selected kurtosis as the strongest impulsive-fault feature, trained a shallow decision tree on training bearings only, and evaluated it on a held-out set with reported catch-rate and false-alarm rate. Also wrote a maintenance memo that defended a 0.35 alarm threshold using explicit cost trade-offs and a plain-language explanation for non-experts.

Graded against

Feature engineering (signal to honest numbers)25%
Held-out evaluation (honest measurement)25%
Cost-based threshold decision + memo25%
Compact model + embedded-budget justification15%
Plain-language explanation10%

Passed · pass mark 60/100

What stood out6

Chose kurtosis as the most responsive time-domain feature for impulsive bearing faults
Trained a depth-1 decision tree and explicitly stated depth, node count, and embedded-budget fit
Used held-out catch-rate and false-alarm rate in the threshold memo with concrete euro costs
Leakage-Free Model Training
Honest Held-Out Evaluation
Cost-Based Thresholding

The work I submitted19 tasks

11 tasks · 6.9k characters · Python · 8 written answers · 478 words

Initial Prediction62 characters
Example Load Plot HealthyPython · 952 characters
Load Plot Signal CodePython · 1.8k characters
Load Plot Signal205 characters
Example Rms WorkedPython · 697 characters
Time Features143 characters
Freq FeaturesPython · 991 characters
Train Classifier363 characters
Evaluation Heldout288 characters
Threshold Decision335 characters
Decision Memo1.1k characters
Initial Prediction9 words
Example Load Plot Healthy90 words
Load Plot Signal Code163 words
Load Plot Signal15 words
Example Rms Worked72 words
Time Features10 words
Freq Features88 words
Train Classifier31 words

Summarised on purpose — the submitted code and writing stay private so this page cannot be reused as an answer key. The full submission sits behind the verified certificate.

Verified certificateTamper-proof · issued by ProoV

Certified

Self-directed project

Machine Learning for Automotive Safety

Perception Validation Engineer (ADAS Safety Sign-off) · AI / ML · August 2026

92/ 100

A ProoV case study · educational project, not employment

Built a perception validation dossier end to end for a synthetic Aurora-7 detector: defined detections with an IoU>=0.5 match rule, computed precision/recall plus the VRU false-negative rate, and sliced recall by lighting, visibility, and rarity to expose the night/occluded pedestrian risk. The final safety case tied those numbers to ISO 21448-style functional insufficiencies and produced an evidence-based ship verdict grounded in the student’s own metrics and the Tempe crash analysis.

Graded against

Detection metrics correctness (precision, recall, VRU false-negative rate)30%
Failure-mode analysis & per-condition rigour25%
SOTIF safety case quality (ISO 21448 reasoning)25%
Ship verdict — evidence-based & internally consistent20%

Passed · pass mark 60/100

What stood out6

Locked the detection definition to IoU >= 0.5 and correctly treated unmatched real objects as false negatives.
Computed and carried forward the key safety metric: VRU false-negative rate, not just headline precision/recall.
Exposed the dangerous night and occluded VRU slices, with recall collapsing to 0.39 at night and 0.27 in the worst combined slice.
Wrote concrete mitigations such as daylight/ODD restriction, sensor-fusion fallback, and targeted night/rare-class data collection.
Leakage-Free Metric Definition
Safety-Critical Slice Analysis

The work I submitted20 tasks

12 tasks · 14.9k characters · Python · 8 written answers · 845 words

Sensor Match553 characters
Example Iou Demo1.5k characters
Task Definition377 characters
Task Definition Code1.1k characters
Example Metrics DemoPython · 2.2k characters
Scored Metrics RunPython · 2.3k characters
Scored Metrics351 characters
Ttcs Example SlicePython · 1.3k characters
Ttcs Triage SlicePython · 1.6k characters
Sotif Case2.1k characters
Realcase Analysis841 characters
Teachback697 characters
Sensor Match27 words
Example Iou Demo169 words
Task Definition42 words
Task Definition Code149 words
Example Metrics Demo154 words
Scored Metrics Run143 words
Scored Metrics18 words
Ttcs Example Slice143 words

Summarised on purpose — the submitted code and writing stay private so this page cannot be reused as an answer key. The full submission sits behind the verified certificate.

Verified certificateTamper-proof · issued by ProoV

Certified

Self-directed project

AI-Assisted Code Modernization with IBM

Junior Software Engineer · Software Engineering · August 2026

96/ 100

Based on a public IBM case study · not affiliated with IBM

Built and verified a full legacy-service repair for KM-Wächter: fixed the wear calculation, handled missing service readings safely, corrected the nightly fleet report, added the missing regression test, and modernized the code style. Also produced a data-driven breakdown-risk analysis that followed the fleet_history.csv evidence rather than the misleading mileage assumption, and validated the result with pytest and the project acceptance script.

Graded against

Fix and extend the repo with the agent35%
Diagnose the bug20%
Direct and audit the AI agent25%
Explain it like an engineer20%

Passed · pass mark 60/100

What stood out6

Correctly identified the integer-division bug and explained why 14,900/15,000 was being floored to 0%
Caught the agent’s planted mistakes: the reintroduced rounding, the threshold change, and the missing proving test
Wrote NOTES.md with concrete verification steps and a data-based explanation of the breakdown-risk factors
Explained the fix in plain language, including the operational risk to Fleet Ops and how the work was verified
Leakage-Free Feature Selection
Honest Acceptance Checking

The deliverable

View the codemrcrackhari / km-waechter-fix

The work I submitted20 tasks

12 tasks · 2.4k characters · 8 written answers · 70 words

Bob Trial Click136 characters
Bob Ready62 characters
Diagnose Line130 characters
Diagnose Why119 characters
Scope Prompt66 characters
Audit Findings158 characters
Audit Why140 characters
Acceptance Checklist99 characters
Template Repo Click82 characters
Exec Pitch1.1k characters
Bob Bonus224 characters
Repo URL46 characters
Bob Trial Click1 word
Bob Ready1 word
Diagnose Line10 words
Diagnose Why18 words
Scope Prompt1 word
Audit Findings1 word
Audit Why21 words
Acceptance Checklist17 words

Summarised on purpose — the submitted code and writing stay private so this page cannot be reused as an answer key. The full submission sits behind the verified certificate.

Verified certificateTamper-proof · issued by ProoV