601 Labs is the lab and portfolio of Ahamed Kamal. The work below is real — anomaly detection, forecasting and recovery planning, journey analytics, and human-in-the-loop GenAI — told as what the problem was, what I did, and what I learned. Client and employer names left out on purpose; the thinking is the point.
Five problems, and the principle each one taught. Charts are illustrative — representative numbers, drawn to show the idea, not real data.
In large, segmented pipelines, abnormal movements hide in plain sight — and naïve alerts either miss them or cry wolf on every wobble.
I evolved the method in stages: from a fixed threshold, to a global Z-score, to a Z-score computed within each Region × Segment peer group, to an incremental deviation — or “jump” — analysis that scores change over a rolling window rather than raw level.
Context is the whole game. A number is only an outlier relative to its own peer group and its own recent trajectory — and a sudden jump often matters far more than a high or low level.
A region or channel is falling behind plan, and leadership needs more than a gloomy forecast — they need a believable way to close the gap.
Seasonality- and recency-weighted forecasting, made robust with winsorized estimates so a few extreme points don’t distort the baseline; a close-the-gap (CGAP) analysis to size the shortfall; then a recovery plan that allocates uplift where it’s actually achievable.
A forecast is only useful if it’s robust to noise and paired with an action. Prediction without a recovery path is just a number; the value is showing the route back to target.
Skilled analysts were losing whole mornings turning finished analysis into decks and emails — the least valuable use of their time.
A human-in-the-loop pipeline (Python, Apps Script, Colab) that pulls the analysis, drafts the narrative with GenAI, and ships formatted reports and emails — with a human reviewing before anything goes out.
Automation earns trust only with a person in the loop. The win isn’t removing people — it’s removing the drudgery so human judgment gets the time it deserves.
Top-line volume told us how many people showed up, but not where in the journey intent was forming — or quietly draining away.
Mapped the multi-stage journey, scored stage-to-stage conversion, and isolated both the high-drop segments and the high-intent paths worth doubling down on.
The biggest gains hide at the transitions, not the totals. Where people drop between stages tells you more than any single stage’s headline number.
As an individual, how far can you get by orchestrating AI rather than hand-coding everything?
Built 601 Labs AI Desk — a multi-model workspace that routes a question across several frontier models and composes the result — and an AI resume-and-interview tool, shipped in an afternoon.
With good orchestration, one person can build in hours what used to take a team days. The scarce skill is no longer typing the code — it’s knowing what to point the model at, and whether to trust what comes back.
The techniques behind the case studies — from robust statistics to applied GenAI.
I'm Ahamed Kamal, a data & AI strategy leader based in Singapore with 15 years across enterprise BI transformation, data architecture, applied AI, forecasting, and executive decision intelligence.
My strength is working from the business problem backwards: framing ambiguous questions with senior stakeholders, choosing the right data, BI, or AI approach, and driving it through design, delivery, and adoption. AI is how I build faster — the value is knowing what's worth building and whether to trust the result.
601 Labs is where I keep the things I build and the things I'm learning. It may grow into something more.