ISHMAM Jahin

Create your portfolio with ProoV
ProoV
DE Verified work portfolio

ProoV Portfolio

ISHMAM Jahin

Mechanical Engineering · University of Technology Malaysia

View the ProoV leaderboard

Projects

Certified

Self-directed project

Amazon Robotics - Black-Friday Fleet Coordination

Software Engineering · September 2026

96/ 100

A ProoV case study · educational project, not employment

Built an end-to-end Amazon Robotics fleet-coordination prototype for Black Friday operations. The work framed the bottleneck as aisle congestion and dock contention, implemented A* plus centralized reservation-table cooperative planning with vertex and edge protections, chose staggered charging and dispersed idle parking, and validated the policy in simulation with zero collisions and a clear throughput trade-off before pitching a scoped FC pilot to leadership.

Graded against

  • Problem framing & operational stakes20%
  • Algorithmic depth - A* and cooperative coordination30%
  • Coordination policy realism - deadlock, charging, idle25%
  • Throughput trade-off analysis15%
  • Executive pitch to leadership10%

Passed · pass mark 55/100

What stood out6
  • Implemented a reservation table that prevents both vertex collisions and head-on edge swaps
  • Selected a priority-watchdog conflict policy and justified it with simulator trade-offs
  • Identified a concave throughput curve and locked onto a specific fleet-size sweet spot
  • Delivered a leadership-ready memo with a deterministic safety story and scoped pilot ask
  • Leakage-Free Multi-Agent Coordination
  • Honest Simulator-Based Evaluation
The work I submitted20 tasks

12 tasks · 8.0k characters · Python · 8 written answers · 197 words

  • Role And Mission115 characters
  • Amazon Robotics Context459 characters
  • Program Overview133 characters
  • Grid World Theory119 characters
  • State Representation Task ValidatorPython · 1.8k characters
  • State Representation Task341 characters
  • Phase1 Recap348 characters
  • Astar Theory124 characters
  • Astar Implementation Cell3.8k characters
  • Astar Solution Correct4 characters
  • Single Unit Challenge468 characters
  • Phase2 Recap374 characters
  • Role And Mission1 word
  • Amazon Robotics Context37 words
  • Program Overview1 word
  • Grid World Theory1 word
  • State Representation Task Validator132 words
  • State Representation Task1 word
  • Phase1 Recap23 words
  • Astar Theory1 word

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

Energy and Renewable: Machine Learning for the Power Grid

Junior Machine Learning Engineer · AI / ML · September 2026

88/ 100

A ProoV case study · educational project, not employment

Built an end-to-end gas-turbine yield predictor using leakage-free sensor inputs, a time-based held-out split, and a baseline-vs-forest comparison. The candidate reported honest held-out performance (about 1.5 MWh MAE, 1.9 MWh RMSE, R² ≈ 0.98), checked residual behavior, and documented practical limits, including single-plant scope and drift monitoring.

Graded against

  • Build and validate the model correctly35%
  • Optimize and evaluate honestly25%
  • Write the Model Card (clarity + honesty)25%
  • Explain it to a non-expert15%

Passed · pass mark 60/100

What stood out6
  • Correctly excluded CO and NOx as downstream leakage features and explained why they are unavailable at prediction time.
  • Used a genuine time-based train/test split instead of a random shuffle, preserving the temporal structure of the data.
  • Compared a LinearRegression baseline against a RandomForest and showed a clear held-out improvement in MAE and R².
  • Read the residuals rather than stopping at headline metrics, and correctly noted the model’s weaker behavior at high loads.
  • Leakage-Free Feature Selection
  • Honest Held-Out Evaluation
The work I submitted20 tasks

12 tasks · 13 lines · Python, JavaScript · 8 written answers · 437 words

  • Problem Frame196 characters
  • Leakage Flags113 characters
  • Train Validate BaselinePython · 1.4k characters
  • Train Validate Forest943 characters
  • Train Validate124 characters
  • Tune Evaluate SetupPython · 1.4k characters
  • Tune Evaluate ResidualsPython · 2 lines
  • Tune Evaluate130 characters
  • Residual Read406 characters
  • Drift Monitor95 characters
  • Model Card1.6k characters
  • Teach BackJavaScript · 965 characters
  • Problem Frame14 words
  • Leakage Flags1 word
  • Train Validate Baseline141 words
  • Train Validate Forest76 words
  • Train Validate1 word
  • Tune Evaluate Setup111 words
  • Tune Evaluate Residuals92 words
  • Tune Evaluate1 word

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

96/ 100

A ProoV case study · educational project, not employment

Built a complete perception-validation dossier for a synthetic Aurora-7 detector: defined detection rigorously with an IoU ≥ 0.5 match rule, computed aggregate and VRU-specific metrics from a scored table, and showed that VRU false negatives—not headline accuracy—drive the safety case. They then sliced recall by operating condition, identified the night+occluded VRU collapse, wrote ISO 21448-style functional-insufficiency mitigations, and issued a consistent conditional deployment recommendation grounded in their own evidence.

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
  • Computed and interpreted IoU correctly, then locked the detection rule at IoU ≥ 0.5 with unmatched real objects counted as false negatives.
  • Used the full synthetic table to derive TP/FP/FN, precision, aggregate recall, and the VRU false-negative rate, explicitly treating VRU recall as the safety metric.
  • Sliced VRU recall by day/night, clear/occluded, and common/rare, and correctly surfaced the night + occluded slice as the worst-case failure mode.
  • Wrote a conditional deployment verdict that cites the dossier’s own metrics and mitigations, and connected the real Uber Tempe crash to the same night-time VRU failure pattern.
  • Leakage-Free Detection Definition
  • Safety-Critical Metric Selection
The work I submitted20 tasks

12 tasks · 17.3k characters · Python · 8 written answers · 885 words

  • Sensor Match553 characters
  • Task Definition Code1.1k characters
  • Task Definition377 characters
  • Scored Metrics RunPython · 2.3k characters
  • Scored Metrics351 characters
  • Sotif Case3.8k characters
  • Realcase Analysis989 characters
  • Teachback1.2k characters
  • Ship Verdict1.8k characters
  • Example Iou Demo1.5k characters
  • Example Metrics DemoPython · 2.2k characters
  • Ttcs Example SlicePython · 1.3k characters
  • Sensor Match27 words
  • Task Definition Code149 words
  • Task Definition42 words
  • Scored Metrics Run143 words
  • Scored Metrics18 words
  • Sotif Case169 words
  • Realcase Analysis142 words
  • Teachback195 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
3Projects completed
3Verified certificates
93Average score
Create your portfolio with ProoV